Updates to our Transparency Center

• We have combined the category of government takedowns received based on local law and based on Terms of Service into one chart/reporting category of Government takedowns received. We made this change to be more accurate in our reporting; the government takedown requests we receive may cite a local law or a Terms of Service violation, but more typically, they are just official requests for the removal of content. 
• We have retained the separate categories of “Government takedowns processed based on local law” and “Government takedowns processed based on Terms of Service.” This is an important distinction because it reflects that some content governments ask us to take down is in violation of our Terms and is processed like any other report, whereas some content is not in violation of our terms, but is in violation of local law. In the latter case, we limit the impact on developers by only restricting access to the content in the jurisdiction in which we are legally required to do so, and we publish the request in our gov-takedowns repository to ensure transparency.
• We have also clarified the README of our gov-takedowns repository to note that the repository solely contains official government requests, which led to content removal based on local law. 

Updates to our Acceptable Use Policies

• Child Sexual Abuse Material (CSAM)
• Terrorist & Violent Extremist Content (TVEC)
• Non-Consensual Intimate Imagery
• Synthetic Media and AI Tools

Updates to our Transparency Center

• We have combined the category of government takedowns received based on local law and based on Terms of Service into one chart/reporting category of Government takedowns received. We made this change to be more accurate in our reporting; the government takedown requests we receive may cite a local law or a Terms of Service violation, but more typically, they are just official requests for the removal of content. 
• We have retained the separate categories of “Government takedowns processed based on local law” and “Government takedowns processed based on Terms of Service.” This is an important distinction because it reflects that some content governments ask us to take down is in violation of our Terms and is processed like any other report, whereas some content is not in violation of our terms, but is in violation of local law. In the latter case, we limit the impact on developers by only restricting access to the content in the jurisdiction in which we are legally required to do so, and we publish the request in our gov-takedowns repository to ensure transparency.
• We have also clarified the README of our gov-takedowns repository to note that the repository solely contains official government requests, which led to content removal based on local law. 

Updates to our Acceptable Use Policies

• Child Sexual Abuse Material (CSAM)
• Terrorist & Violent Extremist Content (TVEC)
• Non-Consensual Intimate Imagery
• Synthetic Media and AI Tools

背景

快速示例

1

grep -n -C 1 "Fatal error" app.log

• -n 会显示行号，便于定位。
• -C 1 等价于 --context=1，表示向前向后各多带 1 行。想多看几行时调整数字即可。

控制上下文范围

• -C <N>：两侧各 N 行，是最常用的形式。
• -B <N>：只带前 N 行（Before）。
• -A <N>：只带后 N 行（After）。

1

grep -n -A 4 "NullPointerException" stacktrace.txt

与常见参数组合

• -i：忽略大小写，处理大小写不一致的告警信息很方便。
• -E：启用扩展正则，可直接写 grep -E "(WARN|ERROR)"。
• --color=auto：高亮命中关键字，在终端阅读更直观。

1
2
3
4

gctx() {
  local keyword="$1" file="$2" lines="${3:-1}"
  grep -n --color=always -C "$lines" "$keyword" "$file"
}

小结

• -C/-A/-B 是获取上下文的核心选项，记住数字表示行数即可。
• 搭配 -n、--color、-m 等参数可以提升排查效率。
• 如果命中结果过多，将命令与 less -R 或 fzf 管道组合，能够在终端中进行二次筛选，让排查体验更顺滑。

Diventare il database moderno più popolare al mondo

Il database ideale per l'AI trasformativa

Abbiamo appena iniziato

Entwicklung zur weltweit beliebtesten modernen Datenbank

Die ideale Datenbank für transformative KI

Wir stehen noch ganz am Anfang

세계에서 가장 인기 있는 현대 데이터베이스가 되다

혁신적인 AI를 위한 이상적인 데이터베이스

이제 시작에 불과합니다

Tornar-se o banco de dados moderno mais popular do mundo.

O banco de dados ideal para IA transformativa

Estamos apenas dando os primeiros passos

Devenir la base de données moderne la plus populaire au monde

La base de données idéale pour une IA transformatrice

Et ce n'est que le début

成为全球最受欢迎的现代数据库

变革性 AI 的理想数据库

我们才刚刚入门。

Convertirse en la base de datos moderna más popular del mundo

La base de datos ideal para la IA transformativa

Apenas estamos comenzando

(a) Transactions relating to purchase or sale of security

            It shall be unlawful for any person, directly or indirectly, by the use of the mails or any means or instrumentality of interstate commerce, or of any facility of any national securities exchange, or for any member of a national securities exchange-

                        (1) For the purpose of creating a false or misleading appearance of active trading in any security registered on a national securities exchange, or a false or misleading appearance with respect to the market for any such security, (A) to effect any transaction in such security which involves no change in the beneficial ownership thereof, or (B) to enter an order or orders for the purchase of such security with the knowledge that an order or orders of substantially the same size, at substantially the same time, and at substantially the same price, for the sale of any such security, has been or will be entered by or for the same or different parties, or (C) to enter any order or orders for the sale of any such security with the knowledge that an order or orders of substantially the same size, at substantially the same time, and at substantially the same price, for the purchase of such security, has been or will be entered by or for the same or different parties.

                        (2) To effect, alone or with one or more other persons, a series of transactions in any security registered on a national securities exchange creating actual or apparent active trading in such security or raising or depressing the price of such security, for the purpose of inducing the purchase or sale of such security by others.

                        (3) If a dealer or broker, or other person selling or offering for sale or purchasing or offering to purchase the security, to induce the purchase or sale of any security registered on a national securities exchange by the circulation or dissemination in the ordinary course of business of information to the effect that the price of any such security will or is likely to rise or fall because of market operations of any one or more persons conducted for the purpose of raising or depressing the prices of such security.

                        (4) If a dealer or broker, or other person selling or offering for sale or purchasing or offering to purchase the security, to make, regarding any security registered on a national securities exchange, for the purpose of inducing the purchase or sale of such security, any statement which was at the time and in the light of the circumstances under which it was made, false or misleading with respect to any material fact, and which he knew or had reasonable ground to believe was so false or misleading.

                        (5) For a consideration, received directly or indirectly from a dealer or broker, or other person selling or offering for sale or purchasing or offering to purchase the security, to induce the purchase or sale of any security registered on a national securities exchange by the circulation or dissemination of information to the effect that the price of any such security will or is likely to rise or fall because of the market operations of any one or more persons conducted for the purpose of raising or depressing the price of such security.

                        (6) To effect either alone or with one or more other persons any series of transactions for the purchase and/or sale of any security registered on a national securities exchange for the purpose of pegging, fixing, or stabilizing the price of such security in contravention of such rules and regulations as the Commission may prescribe as necessary or appropriate in the public interest or for the protection of investors.

(b) Transactions relating to puts, calls, straddles, or options

            It shall be unlawful for any person to effect, by use of any facility of a national securities exchange, in contravention of such rules and regulations as the Commission may prescribe as necessary or appropriate in the public interest or for the protection of investors-

                        (1) any transaction in connection with any security whereby any party to such transaction acquires any put, call, straddle, or other option or privilege of buying the security from or selling the security to another without being bound to do so; or

                        (2) any transaction in connection with any security with relation to which he has, directly or indirectly, any interest in any such put, call, straddle, option, or privilege; or

                        (3) any transaction in any security for the account of any person who he has reason to believe has, and who actually has, directly or indirectly, any interest in any such put, call, straddle, option, or privilege with relation to such security.

(c) Endorsement or guarantee of puts, calls, straddles, or options

            It shall be unlawful for any member of a national securities exchange directly or indirectly to endorse or guarantee the performance of any put, call, straddle, option, or privilege in relation to any security registered on a national securities exchange, in contravention of such rules and regulations as the Commission may prescribe as necessary or appropriate in the public interest or for the protection of investors.

(d) Registered warrant, right, or convertible security not included in "put", "call", "straddle", or "option"

            The terms "put", "call", "straddle", "option", or "privilege" as used in this section shall not include any registered warrant, right, or convertible security.

(e) Persons liable; suits at law or in equity

            Any person who willfully participates in any act or transaction in violation of subsections (a), (b), or (c) of this section, shall be liable to any person who shall purchase or sell any security at a price which was affected by such act or transaction, and the person so injured may sue in law or in equity in any court of competent jurisdiction to recover the damages sustained as a result of any such act or transaction.  In any such suit the court may, in its discretion, require an undertaking for the payment of the costs of such suit, and assess reasonable costs, including reasonable attorneys' fees, against either party litigant.  Every person who becomes liable to make any payment under this subsection may recover contribution as in cases of contract from any person who, if joined in the original suit, would have been liable to make the same payment.  No action shall be maintained to enforce any liability created under this section, unless brought within one year after the discovery of the facts constituting the violation and within three years after such violation.

(f) Subsection (a) not applicable to exempted securities

            The provisions of subsection (a) of this section shall not apply to an exempted security.

(g) Foreign currencies

            Notwithstanding any other provision of law, the Commission shall have the authority to regulate the trading of any put, call, straddle, option, or privilege on any security, certificate of deposit, or group or index of securities (including any interest therein or based on the value thereof), or any put, call, straddle, option, or privilege entered into on a national securities exchange relating to foreign currency (but not, with respect to any of the foregoing, an option on a contract for future delivery).

(h) Limitations on practices that affect market volatility

            It shall be unlawful for any person, by the use of the mails or any means or instrumentality of interstate commerce or of any facility of any national securities exchange, to use or employ any act or practice in connection with the purchase or sale of any equity security in contravention of such rules or regulations as the Commission may adopt, consistent with the public interest, the protection of investors, and the maintenance of fair and orderly markets-

                        (1) to prescribe means reasonably designed to prevent manipulation of price levels of the equity securities market or a substantial segment thereof; and

                        (2) to prohibit or constrain, during periods of extraordinary market volatility, any trading practice in connection with the purchase or sale of equity securities that the Commission determines (A) has previously contributed significantly to extraordinary levels of volatility that have threatened the maintenance of fair and orderly markets; and (B) is reasonably certain to engender such levels of volatility if not prohibited or constrained.

In adopting rules under paragraph (2), the Commission shall, consistent with the purposes of this subsection, minimize the impact on the normal operations of the market and a natural person's freedom to buy or sell any equity security.

Interesting look at a Supermax prison … as awful as you can imagine it. Caged animals go barking mad, and it doesn’t help that they are all psychotic to start. The noise on the cellblock is incredible, the fluorescent lights always on, 23 hours a day alone in the cell … it has to be a fate worse than death. Interviews with some very articulate prisoners whom you are relieved are locked up forever.

The prisoners opened up with Kristi Jacobs… she’s an attractive woman.

93 minute running time, so the perfect length … Amy Irving and her cheekbones and 80s hair, channeling Barbara Hershey from Hannah and Her Sisters (1986) … David Pierce, Fraser’s gay little brother, actually name-checks Annie Hall in one scene … New York City looks as diverse and dirty as ever in the late 1980s … there’s a great scene where an old woman breaks into song with “Some Enchanted Evening” (Hammerstein) in a deli and the customers just sort of take it in stride and enjoy it (though Amy herself is deeply affected by it: “Once you have found him, never let him go.”)

This is a very Jewish movie (directed by Joan Micklin Silver, written by Susan Sandler), though Amy Irving was raised Christian Scientist and Peter Riegert comes from a non-observant Jewish family. It’s also clear that Silver loves the city.

There are cute bits, funny bits, some good writing, but ultimately this movie reflects the confused state of feminism by the late 80s… 33-year-old Amy is single and has a job (at a bookstore, of course) and occasional casual sex with a tall, bearded, married man, but she looks longingly at a baby boy at a bris and [spoilers] ultimately accepts the matchmaking arrangement of her Bubbie with a pickle man and his pickle. What are we supposed to make of this?

It’s a deeply traditional, conventional message in the end. It wasn’t terrible, but I can’t recommend it.

生活

ACTUATE 2025: 23–24 September 2025, SAN FRANCISCO

CoRL 2025: 27–30 September 2025, SEOUL

IEEE Humanoids: 30 September–2 October 2025, SEOUL

World Robot Summit: 10–12 October 2025, OSAKA, JAPAN

IROS 2025: 19–25 October 2025, HANGZHOU, CHINA

A billion dollars is a lot of money. And this is actual money, not just a valuation. but Figure already had a lot of money. So what are they going to be able to do now that they weren’t already doing, I wonder?

[ Figure ]

Robots often succeed in simulation but fail in reality. With PACE, we introduce a systematic approach to sim-to-real transfer.

[ Paper ]

Anthropomorphic robotic hands are essential for robots to learn from humans and operate in human environments. While most designs loosely mimic human hand kinematics and structure, achieving the dexterity and emergent behaviors present in human hands, anthropomorphic design must extend to also match passive compliant properties while simultaneously strictly having kinematic matching. We present ADAPT-Teleop, a system combining a robotic hand with human-matched kinematics, skin, and passive dynamics, along with a robotic arm for intuitive teleoperation.

[ Paper ]

This robot can walk without any electronic components in its body, because the power is transmitted through wires from motors concentrated outside of its body. Also, this robot’s front and rear legs are optimally coupled and can walk with just four wires.

[ JSK Lab ]
Thanks, Takahiro!

Five teams of Los Alamos engineers competed to build the ultimate hole-digging robot dog in a recent engineering sprint. In just days, teams programmed their robot dogs to dig, designing custom “paws” from materials like sheet metal, foam, and 3D-printed polymers. The paws mimicked animal digging behaviors—from paddles and snowshoes to dew claws—and helped the robots avoid sinking into a 30-gallon soil bucket. Teams raced to see whose dog could dig the biggest hole and dig under a fence the fastest.

[ Los Alamos ]

This work presents UniPilot, a compact hardware-software autonomy payload that can be integrated across diverse robot embodiments to enable resilient autonomous operation in GPS-denied environments. The system integrates a multimodal sensing suite including lidar, radar, vision, and inertial sensing for robust operation in conditions where unimodal approaches may fail. A large number of experiments are conducted across diverse environments and on a variety of robot platforms to validate the mapping, planning, and safe navigation capabilities enabled by the payload.

[ NTNU ]
Thanks, Kostas!

KAIST Humanoid v0.5. Developed at the DRCD Lab, KAIST, with a control policy trained via reinforcement learning.

[ KAIST ]

I just like the determined little hops.

[ AgileX ]

I’m always a little bit suspicious of robotics labs that are exceptionally clean and organized.

[ PNDbotics ]

Er, has PAL Robotics ever actually seen a kangaroo...?

[ PAL ]

See Spots push. Push, Spots, push.

[ Tufts ]

Training humanoid robots to hike could accelerate development of embodied AI for tasks like autonomous search and rescue, ecological monitoring in unexplored places, and more, say University of Michigan researchers who developed an AI model that equips humanoids to hit the trails.

[ Michigan ]

I am dangerously close to no longer being impressed by breakdancing humanoid robots.

[ Fourier ]

This, though, would impress me.

[ Inria ]

In this interview, Clone’s co-founder and CEO Dhanush Radhakrishnan discusses the company’s path to creating the synthetic humans straight out of science fiction.
(If YouTube brilliantly attempts to auto-dub this for you, switch the audio track to original [which YouTube thinks is Polish] and the video will still be in English.)

[ Clone ]

This documentary takes you behind the scenes of the HMND 01 Alpha release: the breakthroughs, the failures, and the late nights of building the U.K.’s first industrial humanoid robot.

[ Humanoid ]

What is the role of ethical considerations in the development and deployment of robotic and automation technologies, and what are the responsibilities of researchers to ensure that these technologies advance in ways that are transparent, fair, and aligned with the broader well-being of society?

[ ICRA@40 ]

This UPenn GRASP SFI lecture is from Tairan He at Nvidia on “Scalable Sim-to-Real Learning for General-Purpose Humanoid Skills.”

Humanoids represent the most versatile robotic platform, capable of walking, manipulating, and collaborating with people in human-centered environments. Yet despite recent advances, building humanoids that can operate reliably in the real world remains a fundamental challenge. Progress has been hindered by difficulties in whole-body control, robust perceptive reasoning, and bridging the sim-to-real gap. In this talk, I will discuss how scalable simulation and learning can systematically overcome these barriers.
[ UPenn ]

How Molten Regolith Electrolysis Reactors Work

Are You Smarter Than an AI?

Why It’s So Hard to Test for Intelligence

Clever Hans, a German horse in the early 1900s, seemed able to do math—but was really responding to his trainer’s subtle cues, a classic case of misinterpreting performance. Alamy
Intelligence testing is hard—in people, animals, or machines. You must beware of both false positives and false negatives. Maybe the test taker appears smart only by taking shortcuts, like Clever Hans, the famous horse that appeared to be capable of math but actually responded to nonverbal cues. Or maybe test takers appear stupid only because they are unfamiliar with the testing procedure or have perceptual difficulties.
It’s also hard because notions of intelligence vary across place and time. “There is an interesting shift in our society in terms of what we think intelligence is and what aspects of it are valuable,” says Anna Ivanova, an assistant professor of psychology at Georgia Tech. For example, before encyclopedias and the Internet, “having a large access to facts in your head was considered a hallmark of intelligence.” Now we increasingly prize fluid over crystallized intelligence.

The History of AI Intelligence Tests

IBM’s Deep Blue defeated world chess champion Garry Kasparov in 1997, butdidn’t have enough general intelligence to play checkers. Adam Nadel/AP
In 1950, Alan Turing proposed the imitation game, a version of which requires a machine to pass as a human in typewritten conversation. “The question and answer method seems to be suitable for introducing almost any one of the fields of human endeavour that we wish to include,” he wrote. For decades, passing what’s now called the Turing test was considered a nearly impossible challenge and a strong indicator of AGI.
But this year, researchers reported that when people conversed with both another person and OpenAI’s GPT-4.5 for 5 minutes and then had to guess which one was human, they picked the AI 73 percent of the time. Meanwhile, top language models frequently make mistakes that few people ever would, like miscounting the number of times the letter r occurs in strawberry. They appear to be more wheel than human leg. So scientists are still searching for measures of humanlike intelligence that can’t be hacked.

The ARC Test for AGI

The Abstraction and Reasoning Corpus challenges AI systems to infer abstract rules from just a few examples. Given examples of input-output grids, the system must apply the hidden pattern to a new test case—something humans find easy but machines still struggle with. ARC Prize
ARC focuses on fluid intelligence. “To solve any problem, you need some knowledge, and then you’re going to recombine that knowledge on the fly,” Chollet told me. To make it a test not of stored knowledge but of how one recombines it, the training puzzles are supposed to supply all the “core knowledge priors” one needs. These include concepts like object cohesion, symmetry, and counting—the kind of common sense a small child has. Given this training and just a few examples, can you figure out which knowledge to apply to a new puzzle? Humans can do most of the puzzles easily, but AI struggled, at least at first. Eventually, OpenAI created a version of its o3 reasoning model that outperformed the average human test taker, achieving a score of 88 percent—albeit at an estimated computing cost of US $20,000 per puzzle. (OpenAI never released that model, so it’s not on the leaderboard chart.)
This March, Chollet introduced a harder version, called ARC-AGI-2. It’s overseen by his new nonprofit, the ARC Prize Foundation. “Our mission is to serve as a North Star towards AGI through enduring benchmarks,” the group announced. ARC Prize is offering a million dollars in prize money, the bulk going to teams whose trained AIs can solve 85 percent of 120 new puzzles using only four graphics processors for 12 hours or less. The new puzzles are more complex than those from 2019, sometimes requiring the application of multiple rules, reasoning for multiple steps, or interpreting symbols. The average human score is 60 percent, and as of this writing the best AI score is about 16 percent.

AI models have made gradual progress on the first version of the ARC-AGI benchmark, which was introduced in 2019. This year, the ARC Prize launched a new version with harder puzzles, which AI models are struggling with. Models are labeled low, medium, high, or thinking to indicate how much computing power they expend on their answers, with “thinking” models using the most.ARC Prize
AI experts acknowledge ARC’s value, and also its flaws. Jiaxuan You, a computer scientist at the University of Illinois at Urbana-Champaign, says ARC is “a very good theoretical benchmark” that can shed light on how algorithms function, but “it’s not taking into account the real-world complexity of AI applications, such as social reasoning tasks.”
Melanie Mitchell, a computer scientist at the Santa Fe Institute, says it “captures some interesting capabilities that humans have,” such as the ability to abstract a new rule from a few examples. But given the narrow task format, she says, “I don’t think it captures what people mean when they say general intelligence.”
Despite these caveats, ARC-AGI-2 may be the AI benchmark with the biggest performance gap between advanced AI and regular people, making it a potent indicator of AGI’s headway. What’s more, ARC is a work in progress. Chollet says AI might match human performance on the current test in a year or two, and he’s already working on ARC-AGI-3. Each task will be like a miniature video game, in which the player needs to figure out the relevant concepts, the possible actions, and the goal.

What Attributes Should an AGI Benchmark Test?

Virtual people would be assigned randomized tasks that test not only understanding but values. For example, AIs might unexpectedly encounter money on the floor or a crying baby.
One 2024 paper in Engineering proposed the Tong test (tong is Chinese for “general”). Virtual people would be assigned randomized tasks that test not only understanding but values. For example, AIs might unexpectedly encounter money on the floor or a crying baby, giving researchers the opportunity to observe what the AIs do. The authors argue that benchmarks should test an AI’s ability to explore and set its own goals, its alignment with human values, its causal understanding, and its ability to control a virtual or physical body. What’s more, the benchmark should be capable of generating an infinite number of tasks involving dynamic physical and social interactions.
Others, like Minsky, have suggested tests that require interacting with the real world to various degrees: making coffee in an unfamiliar kitchen, turning a hundred thousand dollars into a million, or attending college on campus and earning a degree. Unfortunately, some of these tests are impractical and risk causing real-world harm. For example, an AI might earn its million by scamming people.
I asked Hinton, the Nobel Prize winner, what skills will be the hardest for AI to acquire. “I used to think it was things like figuring out what other people are thinking,” he said, “but it’s already doing some of that. It’s already able to do deception.” (In a recent multi-university study, an LLM outperformed humans at persuading test takers to select wrong answers.) He went on: “So, right now my answer is plumbing. Plumbing in an old house requires reaching into funny crevices and screwing things the right way. And I think that’s probably safe for another 10 years.”
Researchers debate whether the ability to perform physical tasks is required to demonstrate AGI. A paper from Google DeepMind on measuring levels of AGI says no, arguing that intelligence can show itself in software alone. They frame physical ability as an add-on rather than a requirement for AGI.
Mitchell of the Santa Fe Institute says we should test capabilities involved in doing an entire job. She noted that AI can do many tasks of a human radiologist but can’t replace the human because the job entails a lot of tasks that even the radiologist doesn’t realize they’re doing, like figuring out what tasks to do and dealing with unexpected problems. “There’s such a long tail of things that can happen in the world,” she says. Some robotic vacuum cleaners weren’t trained to recognize dog poop, she notes, and so they smeared it around the carpet. “There’s all kinds of stuff like that that you don’t think of when you’re building an intelligent system.”
Some scientists say we should observe not only performance but what’s happening under the hood. A recent paper coauthored by Jeff Clune, a computer scientist at the University of British Columbia, in Canada, reports that deep learning often leads AI systems to create “fractured entangled representations”—basically a bunch of jury-rigged shortcuts wired together. Humans, though, look for broad, elegant regularities in the world. An AI system might appear intelligent based on one test, but if you don’t know the system’s innards, you could be surprised when you deploy it in a new situation and it applies the wrong rule.

AGI Is Already Here, and Never Will Be

If you want to get a sense of the truly global scale of the electronics industry, look no further than your smartphone. The processor that powers it started as a humble rock, and by the time it found its way into your device, it had probably seen more of the world than you have. Along the way it was subjected to some of the most technologically sophisticated and closely guarded processes on the planet. Come along as we retrace that incredible 30,000-kilometer ride.

1. Quartz

2. Silicon Metal

3. Purified Polysilicon

4. Silicon Wafers

5. Processed Wafers

6. Packaged Chips

7. Smartphones

Global Trade

Acknowledgment: This journey was inspired by a chapter in Ed Conway’s Material World: The Six Raw Materials That Shape Modern Civilization (Alfred A. Knopf, 2023).
This article appears in the October 2025 print issue as “From Silica to Smartphone.”

Google

Stack Overflow

IEEE Xplore Digital Library

IEEE Jobs Site

Career Builder

GitHub

Trinity College Dublin Library

Discord

What’s the future of programming languages?

Vacuum measurement on Earth

A portable cold-atom vacuum-measurement tool [top] detects the fluorescence of roughly 1 million lithium atoms [bottom], and infers the vacuum pressure based on how quickly the fluorescence decays. Photos: Jayme Thornton

Indeed, an ionization gauge has the same basic components as a vacuum tube. The gauge contains a heated filament that emits electrons into the chamber. The electrons are accelerated toward a positively charged grid. En route to the grid, the electrons occasionally collide with atoms and molecules flying around in the vacuum, knocking off their electrons and creating positively charged ions. These ions are then collected by a negatively charged electrode. The current generated by these positive ions is proportional to the number of atoms floating about in the vacuum, giving a pressure reading.
Ion gauges are relatively cheap (under US $1,000) and commonplace. However, they come with a few difficulties. First, although the current in the ion gauge is proportional to the pressure in the chamber, that proportionality constant depends on a lot of fine details, such as the precise geometry of the filament and the grid. The current cannot be easily calculated from the electrical and physical characteristics of the setup—ion gauges require thorough calibrations. “A full calibration run on the ion gauges is like a full month of somebody’s time,” says Daniel Barker, a physicist at NIST who’s also working on the cold-atom vacuum-measurement project.
Second, the calibration services provided by NIST (among others) calibrate down to only 10^-7 Pa. Performance below that pressure is questionable, even for a well-calibrated gauge. What’s more, at lower pressures, the heat from the ion gauge becomes a problem: Hotter surfaces emit atoms in a process called outgassing, which pollutes the vacuum. “If you’re shooting for a vacuum chamber with really low pressures,” Madison says, “these ionization gauges actually work against you, and many people turn them off.”
Third, the reading on the ion gauge depends very strongly on the types of atoms or molecules present in the vacuum. Different types of atoms could produce readings that vary by up to a factor of four. This variance is fine if you know exactly what’s inside your vacuum chamber, or if you don’t need that precise a measurement. But for certain applications, especially in research settings, these concerns are significant.

How a cold-atom vacuum standard works

The most common vacuum-measurement tool in the high-vacuum range today is the ion gauge, basically a vacuum tube in reverse: A hot filament emits electrons that fly toward a positively charged grid, ionizing background atoms and molecules along the way. Jayme Thornton
UBC’s Kirk Madison and his collaborator James Booth (then at the British Columbia Institute of Technology, in Burnaby), were among the first to turn that thinking on its head back in the 2000s. Instead of battling the vacuum to preserve the trapped atoms, they thought, why not use the trapped atoms as a sensor to measure how empty the vacuum is?To understand how they did that, consider a typical cold-atom vacuum-measurement device. Its main component is a vacuum chamber filled with a vapor of a particular atomic species. Some experiments use rubidium, while others use lithium. Let’s call it lithium between friends.
A tiny amount of lithium gas is introduced into the vacuum, and some of it is captured in a magneto-optical trap. The trap consists of a magnetic field with zero intensity at the center of the trap, increasing gradually away from the center. Six laser beams point toward the center from above, below, the left, the right, the front, and the back. The magnetic and laser forces are arranged so that any lithium atom that might otherwise fly away from the center is most likely to absorb a photon from the lasers, getting a momentum kick back into the trap.
The trap is quite shallow, meaning that hot atoms—above 1 kelvin or so—will not be captured. So the result is a small, confined cloud of really cold atoms, at the center of the trap. Because the atoms absorb laser light occasionally to keep them in the trap, they also reemit light, creating fluorescence. Measuring this fluorescence allows scientists to calculate how many atoms are in the trap.
To use this setup to measure vacuum, you load the atoms into the magneto-optical trap and measure the fluorescence. Then, you turn off the light and hold the atoms in just the magnetic field. During this time, background atoms in the vacuum will chance upon the trapped atoms, knocking them out. After a little while, you turn the light back on and check how much the fluorescence has decreased. This measures how many atoms got knocked out, and therefore how many collisions occurred.
The reason you need the trap to be so shallow and the atoms to be so cold is that these collisions are very weak. “A few collisions are quite energetic, but most of the background gas particles fly by and, like, whisper to the trapped atom, and it just gently moves away,” Madison says.
This method has several advantages over the traditional ion-gauge measurement. The atomic method does not need calibration; the rate at which fluorescence dims depending on the vacuum pressure can be calculated accurately. These calculations are involved, but in a paper published in 2023 the NIST team demonstrated that the latest method of calculation shows excellent agreement with their experiment. Because this technique does not require calibration, it can serve as a primary standard for vacuum pressure, and even potentially be used to calibrate ion gauges.
The cold-atom measurement is also much less finicky when it comes to the actual contents of the vacuum. Whether the vacuum is contaminated with helium or plutonium, the measured pressure will vary by perhaps only a few percent, while the ion gauge sensitivity and reading for these particles might differ by an order of magnitude, Eckel says.
Cold atoms could also potentially measure much lower vacuum pressures than ion gauges can. The current lowest pressure they’ve reliably measured is around 10^-9 Pa, and NIST scientists are working on figuring out what the lower boundary might be. “We honestly don’t know what the lower limit is, and we’re still exploring that question,” Eckel says.

No vacuum is completely empty. The degree to which vacuum pressure approaches pure nothingness is measured in pascals, with Earth’s atmosphere clocking in at 10⁵ Pa and intergalactic space at a measly 10^-20. In between, the new cold-atom vacuum gauges can measure further along the emptiness scale than the well-established ionization gauges can.

Sources: S. Eckel (cold-atom gauge, ionization gauge); K. Zou (molecular-beam epitaxy, chemical vapor deposition); L. Monteiro, “1976 Standard Atmosphere Properties” (Earth’s atmosphere); E.J. Öpik, Planetary and Space Science (1962) (Mars, moon atmosphere); A. Chambers, ‘Modern Vacuum Physics” (2004) (interplanetary and intergalactic space)
Of course, the cold-atom approach also has drawbacks. It struggles at higher pressure, above 10^-7 Pa, so its applications are confined to the ultrahigh vacuum range. And, although there are no commercial atomic vacuum sensors available yet, they are likely to be much more expensive than ion gauges, at least to start.
That said, there are many applications where these devices could unlock new possibilities. At large science experiments, including LIGO (the Laser Interferometer Gravitational-Wave Observatory) and ones at CERN (the European Organization for Nuclear Research), well-placed cold-atom vacuum sensors could measure the vacuum pressure and also help determine where a potential leak might be coming from.
In semiconductor development, a particularly promising application is molecular-beam epitaxy (MBE). MBE is used to produce the few, highly pure semiconductor layers used in laser diodes and devices for high-frequency electronics and quantum technologies. The technique functions in ultrahigh vacuum, with pure elements in separate containers heated on one end of the vacuum. The elements travel across the vacuum until they hit the target surface, where they grow one layer at a time.
Precisely controlling the proportion of the ingredient elements is essential to the success of MBE. Normally, this requires a lot of trial and error, building up thin films and checking whether the proportions are correct, then adjusting as needed. With a cold-atom vacuum sensor, the quantity of each element emitted into the vacuum can be detected on the fly, greatly speeding up the process.
“If this technique could be used in molecular-beam epitaxy or other ultrahigh vacuum environments, I think it will really benefit materials development,” says Ke Zou, an assistant professor of physics at UBC who studies molecular-beam epitaxy. In these high-tech industries, researchers may find that the ability to measure nothing is everything.
This article appears in the October 2025 print issue.

These charts show how employment over time compares among early career, developing, and senior workers (all occupations). Each age group is divided into five groups, based on AI exposure, and normalized to 1 in October 2022—roughly when popular generative AI tools became available to the public.
The trend may be a harbinger for more widespread changes, and the researchers plan to continue tracking the data. “It could be that there are reversals in these employment declines. It could be that other age groups become more or less exposed [to generative AI] and have differing patterns in their employment trends. So we’re going to continue to track this and see what happens,” says Bharat Chandar, one of the paper’s authors and a postdoctoral fellow at the Stanford Digital Economy Lab. In the most AI “exposed” jobs, AI tools can assist with or perform more of the work people do on a daily basis.
So, what does this mean for engineers?

Software Developers Among Most AI-Exposed

Since late 2022, employment for early-career software developers has dropped. Employment for other age groups, however, has seen modest growth.
Chandar cautions that, for specific occupations, the trend may not be driven by AI alone; other changes in the tech industry could also be causing the drop. Still, the fact that it holds across industries suggests that there’s a real effect from AI.
The Stanford team also looked at a broader category of “computer occupations” based on the U.S. Bureau of Labor classifications—which includes hardware engineers, web developers, and more—and found similar results.

Growth in employment between October 2022 and July 2025 by age and AI exposure group. Quintiles 1-3 represent the lowest AI exposure groups, which experienced 6-13 percent growth. Quintiles 4-5 are the most AI-exposed jobs; employment for the youngest workers in these jobs fell 6 percent.

Augmentation vs. Automation

Developing Microcredentials

A student at UC Santa Barbara loads wafers into a machine used for plasma etching. Ben Werner
Since then, he has collaborated with several other institutions and organizations working to implement a microcredential program developed by IEEE in partnership with the University of Southern California (USC) as part of California DREAMS (Defense Ready Electronics and Microdevices Superhub), funded by the DOD. Other programs also offer short training courses, but the standardization IEEE aims to provide is important for ensuring participants’ skills are widely recognized by employers across the country.
Initially, John aimed to address the shortage of technicians to help companies scale up production. But as the courses have expanded elsewhere, it has become clear that the same hands-on experience can be used for engineering students as well.
Students who take these introductory courses may go on to join the workforce or continue in their education to a bachelor’s or advanced degree. “The entire ladder of different workforce exits into the semiconductor industry is really important,” says John. The industry needs operators and technicians, who may seek employment right after high school, as well as Ph.D.-level engineers. “These microcredentials get somebody into the start of that workforce ladder.”

What the Semiconductor Industry Needs

“It doesn’t matter exactly which specific device you made. What matters is that this person has had the experience of making some real chip.” —Demis John
The course has students spend several days in a clean room, which is more likely than a single clean-room day to filter out participants who wouldn’t last. That’s important for companies that invest a lot of resources in hiring and training new people, notes the University of Washington’s Darick Baker, who serves as acting director of the Washington Nanofabrication Facility, in Seattle.

Can Hands-On Courses Scale Up?

Unitree Robots’ BLE Security Flaw Exposed

How to Secure Unitree Robots

CoRL 2025: 27–30 September 2025, SEOUL

IEEE Humanoids: 30 September–2 October 2025, SEOUL

World Robot Summit: 10–12 October 2025, OSAKA, JAPAN

IROS 2025: 19–25 October 2025, HANGZHOU, CHINA

Gemini Robotics 1.5 is our most capable vision-language-action (VLA) model that turns visual information and instructions into motor commands for a robot to perform a task. This model thinks before taking action and shows its process, helping robots assess and complete complex tasks more transparently. It also learns across embodiments, accelerating skill learning.

[ Google DeepMind ]

A simple “force pull” gesture brings Carter straight into her hand. This is a fantastic example of how an intuitive interaction can transform complex technology into an extension of our intent.

[ Robust.ai ]

I can’t help it, I feel bad for this poor little robot.

[ Urban Robotics Laboratory, KAIST ]

Hey look, no legs!

[ Kinisi Robotics ]

Researchers at the University of Michigan and Shanghai Jiao Tong University have developed a soft robot that can crawl along a flat path and climb up vertical surfaces using its unique origami structure. The robot can move with an accuracy typically seen only in rigid robots.

[ University of Michigan Robotics ]

Unitree G1 has learned the “Anti-Gravity” mode: stability is greatly improved under any action sequence, and even if it falls, it can quickly get back up.

[ Unitree ]

Kepler Robotics has commenced mass production of the K2 Bumblebee, the world’s first commercially available humanoid robot powered by Tesla’s hybrid architecture.

[ Kepler Robotics ]

Reinforcement learning (RL)-based legged locomotion controllers often require meticulous reward tuning to track velocities or goal positions while preserving smooth motion on various terrains. Motion imitation methods via RL using demonstration data reduce reward engineering but fail to generalize to novel environments. We address this by proposing a hierarchical RL framework in which a low-level policy is first pre-trained to imitate animal motions on flat ground, thereby establishing motion priors. Real-world experiments with an ANYmal-D quadruped robot confirm our policy’s capability to generalize animal-like locomotion skills to complex terrains, demonstrating smooth and efficient locomotion and local navigation performance amidst challenging terrains with obstacles.

[ ETHZ RSL ]

I think we have entered the ‘differentiation-through-novelty’ phase of robot vacuums.

[ Roborock ]

In this work, we present Kinethreads: a new full-body haptic exosuit design built around string-based motor-pulley mechanisms, which keeps our suit lightweight (<5kg), soft and flexible, quick-to-wear (<30 seconds), comparatively low-cost (~$400), and yet capable of rendering expressive, distributed, and forceful (up to 120N) effects.

[ ACM Symposium on User Interface and Software Technology ]

In this episode of the IBM AI in Action podcast, Aaron Saunders, CTO of Boston Dynamics, delves into the transformative potential of AI-powered robotics, highlighting how robots are becoming safer, more cost-effective and widely accessible through Robotics as a Service (RaaS).

[ IBM ]

This CMU RI Seminar is by Michael T. Tolley from UCSD, on ‘Biologically Inspired Soft Robotics.’

Robotics has the potential to address many of today’s pressing problems in fields ranging from healthcare to manufacturing to disaster relief. However, the traditional approaches used on the factory floor do not perform well in unstructured environments. The key to solving many of these challenges is to explore new, non-traditional designs. Fortunately, nature surrounds us with examples of novel ways to navigate and interact with the real world. Dr. Tolley’s Bioinspired Robotics and Design Lab seeks to borrow the key principles of operation from biological systems and apply them to robotic design.
[ Carnegie Mellon University Robotics Institute ]

Building the world’s most sensitive antenna

Langley engineers (from right): Norman Crabill, Edwin Kilgore, and an unidentified man take a peek inside the vast balloon during inflation tests of the Echo 1 Satellite in Weeksville, N.C.NASA
To make the communication possible, project leads developed and built the horn antenna on the Bell Labs site. The antenna was 15.24 meters long by 6.1 meters wide, weighing in at 16,329 kilograms. It funneled radio waves in or out of the horn shape, and the reflector bounced the waves into a single focused beam—similar to a huge metal megaphone pointing into a curved mirror. Despite its large size, the machine could be precisely aimed.
Unlike other antennas that are tuned to only one frequency, the Holmdel antenna worked across a wide band of frequencies, so it could pick up several types of radio signals. It also could handle radio waves moving in linear or circular paths.
The design accounted for the potential need to eliminate unwanted noise from the environment.
The receiver was placed at the horn’s apex, eliminating the need for a connecting line, which could result in external noise and signal loss.
The antenna allowed Project Echo to complete the first high-quality long-distance voice circuit in 1961 through its namesake’s passive communication satellite, Echo. A similar experiment was successfully completed two years later through the Telstar satellite, according to the proposal for the IEEE Milestone.
In 1964 Penzias and Wilson began using the Holmdel antenna to perform their own radio astronomy experiments.

What’s that buzzing sound?

Penzias and Wilson’s evidence for the big bang theory shaped “our understanding of this universe and our place in it.” —Thomas Coughlin, 2024 IEEE president
“They went so far as to take rags and detergents to carefully wash the antenna from the droppings of a pair of pigeons that had nested there,” Leonardo Colletti told IEEE Spectrum in a 2023 article about the discovery. Colletti is a physics professor at the Free University of Bozen-Bolzano, in Italy.
But even after all the duo’s work, the mysterious buzz continued.
After Penzias and Wilson had accounted for everything, including the pigeon poop, they concluded that the radiation they detected could not have come from the Earth, the sun, or anything else in the galaxy.
They later learned that researchers and astrophysicists Robert H. Dicke, P. James Peebles, and David Todd Wilkinson at Princeton University predicted the existence of cosmic microwave background noise, which “they believed would have resulted from the big bang,” according to an entry on the Nokia Bell Labs website.
“As it turned out,” the article says, “the radiation detected by Penzias and Wilson was a perfect match for what the Princeton researchers had predicted.”

Saving the horn antenna

[From left] Holmdel Deputy Mayor Kim LaMountain, former Bell Labs researcher Giovanni Vannucci, and 2024 IEEE President Tom Coughlin celebrating the Milestone dedication in front of the Horn Antenna in Holmdel, N.J.Bala Prasanna
That put the landmark at risk of being demolished, IEEE Spectrum reported.
The local community banded together, launching a publicity campaign and an online petition to save the antenna. The township ultimately secured ownership of the horn antenna site following an extensive legal process. Last year it dedicated the site as Dr. Robert Wilson Park, honoring it as the place where “we gained a critical understanding of the birth of our universe.”
A plaque recognizing the IEEE Milestone designation is displayed in the lobby of the AT&T Labs Science and Technology Center in Middletown, N.J., which is about 7 kilometers from Crawford Hill. The plaque reads:
In 1959–1960, NASA and AT&T developed a satellite Earth station in Holmdel, N.J., including a novel tracking horn-reflector antenna, maser preamplifier, and FM demodulator. The Earth station demonstrated the first high-quality long-distance voice circuit via the Echo passive communication satellite in 1960–1961, and via the active Telstar communications satellite in 1962–1963. Experiments conducted in 1964–1965 provided the first indication of the cosmic background radiation associated with the Big Bang.
The IEEE New Jersey Coast Section and the IEEE Photonics Society sponsored the nomination.
Administered by the IEEE History Center and supported by donors, the Milestone program recognizes outstanding technical developments around the world.

启动 MinIO

1
2
3
4
5
6
7
8

docker run -d --rm --name minio \
  -e MINIO_ROOT_USER=miniouser \
  -e MINIO_ROOT_PASSWORD=miniopassword \
  -p 9001:9001 \
  -p 9000:9000 \
  --entrypoint sh \
  minio/minio:latest \
  -c 'mkdir -p /minio_data/starrocks && minio server /minio_data --console-address ":9001"'

1
2

docker exec -it minio sh
mc alias set myminio http://10.0.9.20:9000 miniouser miniopassword; mc admin user svcacct add --access-key AAAAAAAAAAAAAAAAAAAA --secret-key BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB myminio miniouser

启动 cn:

1

docker run -p 9060:9060 -p 8040:8040 -p 9050:9050 -p 8060:8060 -p 9070:9070 -itd --rm --name cn -e "TZ=Asia/Shanghai" starrocks/cn-ubuntu:3.4-latest

1
2

cd cn/config/
echo "priority_networks = 10.0.9.20/24" >> cn.properties

1

bin/start_cn.sh --daemon

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

LOG_DIR = ${STARROCKS_HOME}/log  
  
DATE = "$(date +%Y%m%d-%H%M%S)"  
  
sys_log_level = INFO  
  
http_port = 8030  
rpc_port = 9020  
query_port = 9030  
edit_log_port = 9010  
mysql_service_nio_enabled = true  
  
run_mode = shared_data  
cloud_native_storage_type = S3  
aws_s3_endpoint = 10.0.9.20:9000  
# set the path in MinIO  
aws_s3_path = starrocks  
# credentials for MinIO object read/write  
# 这里的 key 为刚才设置的 access token
aws_s3_access_key = AAAAAAAAAAAAAAAAAAAA  
aws_s3_secret_key = BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB  
aws_s3_use_instance_profile = false  
aws_s3_use_aws_sdk_default_behavior = false  
# Set this to false if you do not want default  
# storage created in the object storage using  
# the details provided above  
enable_load_volume_from_conf = true  

# 本机 IP，需要与 cn 中的配置对齐
priority_networks = 10.0.9.20/24

添加 CN 节点

1
2

ALTER SYSTEM ADD COMPUTE NODE "127.0.0.1:9050";
show compute nodes;

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

CREATE TABLE IF NOT EXISTS par_tbl1
(
    datekey DATETIME,
    k1      INT,
    item_id STRING,
    v2      INT
)PRIMARY KEY (`datekey`,`k1`)
 PARTITION BY date_trunc('day', `datekey`)
 PROPERTIES (
"compression" = "LZ4",
"datacache.enable" = "true",
"enable_async_write_back" = "false",
"enable_persistent_index" = "true",
"persistent_index_type" = "LOCAL",
"replication_num" = "1",
"storage_volume" = "builtin_storage_volume"
);

创建表超时

1

admin set frontend config("tablet_create_timeout_second"="50")

unknown compression type(0) backend [id=10002]

• https://github.com/StarRocks/starrocks/issues/55301
• https://docs.starrocks.io/zh/docs/deployment/shared_data/minio/

1

docker run -p 9060:9060 -p 8040:8040 -p 9050:9050 -p 8060:8060 -p 9070:9070 -itd --rm --name cn -e "TZ=Asia/Shanghai" starrocks/cn-ubuntu:3.5.2

1

docker run -p 9061:9060 -p 8041:8040 -p 9051:9050 -p 8061:8060 -p 9071:9070 -itd --rm --name cn2 -e "TZ=Asia/Shanghai" starrocks/cn-ubuntu:3.5.2

1
2
3

ALTER SYSTEM ADD COMPUTE NODE "127.0.0.1:9050";  
ALTER SYSTEM ADD COMPUTE NODE "127.0.0.1:9051";  
show compute nodes;

经过 debug 发现是在添加节点的时候，由于生成的 workerIpPort 与上一个节点相同（127.0.0.1:9060) 从而导致这个节点被跳过了。

• 再找一个台机器来跑 CN2 节点
• 启动一个虚拟机来跑 CN2 节点
• 使用 docker compose 来启动 CN 集群，会在集群内自动分配不同的 IP
• 利用 Docker Bridge 创建一个虚拟网络，由他来分配 IP

使用 Docker Bridge 网络

1

docker network create --subnet=172.18.0.0/16 --gateway=172.18.0.1 my_custom_net

• --subnet=172.18.0.0/16: 定义网络的 IP 地址范围。这里我们使用了 172.18.x.x 这个私有网段。
• --gateway=172.18.0.1: 指定这个网络的网关地址。

1
2
3

docker run --ip 172.18.0.20 --net my_custom_net -p 9060:9060 -p 8040:8040 -p 9050:9050 -p 8060:8060 -p 9070:9070 -itd --rm --name cn -e "TZ=Asia/Shanghai" starrocks/cn-ubuntu:3.5.2

docker run --ip 172.18.0.30 --net my_custom_net -p 9061:9060 -p 8041:8040 -p 9051:9050 -p 8061:8060 -p 9071:9070 -itd --rm --name cn2 -e "TZ=Asia/Shanghai" starrocks/cn-ubuntu:3.5.2

1
2
3
4
5
6
7
8
9
10

docker run -d --rm --name minio \
  --ip 172.18.0.10 \
  --net my_custom_net \
  -e MINIO_ROOT_USER=miniouser \
  -e MINIO_ROOT_PASSWORD=miniopassword \
  -p 9001:9001 \
  -p 9000:9000 \
  --entrypoint sh \
  minio/minio:latest \
  -c 'mkdir -p /minio_data/starrocks && minio server /minio_data --console-address ":9001"'

1

mc alias set myminio http://172.18.0.10:9000 miniouser miniopassword; mc admin user svcacct add --access-key AAAAAAAAAAAAAAAAAAAA --secret-key BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB myminio miniouser

1
2

ALTER SYSTEM ADD COMPUTE NODE "172.18.0.20:9050";
ALTER SYSTEM ADD COMPUTE NODE "172.18.0.30:9050"

The Blazor Component Model

The Integration Between Blazor and Desktop

Why Desktop Still Matters—and Why Blazor Helps

• Local performance and fewer server resources to manage
• Access to native features
• Offline support
• Offline updates

Reusing Existing Blazor Investments

Common Pitfalls or Gotchas

When to Choose This Path

Conclusion

Prerequisites

What Is Observability?

• Traces: Document the entire flow of a request through the chain of services
• Metrics: Give insights into system performance and resource usage
• Logs: Provide context-rich information that complements traces and metrics

• A single request can interact with multiple layers.
• If X depends on Y, errors or slowness in Y can lead to X failing.
• Logs alone aren’t enough to fully understand what’s happening most of the time.

Understanding OpenTelemetry

Project Setup

nest new Otel-project
cd Otel-project

services:
  jaeger:
    image: jaegertracing/all-in-one:latest
    ports:
      - "16686:16686"
      - "4317:4317"

docker compose up

npm install @nestjs/common @nestjs/core @nestjs/platform-express rxjs \
@opentelemetry/api @opentelemetry/sdk-node @opentelemetry/auto-instrumentations-node \
@opentelemetry/exporter-trace-otlp-grpc @opentelemetry/sdk-metrics

import { NodeSDK } from "@opentelemetry/sdk-node";
import { getNodeAutoInstrumentations } from "@opentelemetry/auto-instrumentations-node";
import { OTLPTraceExporter } from "@opentelemetry/exporter-trace-otlp-grpc";
import {
  PeriodicExportingMetricReader,
  ConsoleMetricExporter,
} from "@opentelemetry/sdk-metrics";

const traceExporter = new OTLPTraceExporter({
  url: "http://localhost:4317",
});

const metricReader = new PeriodicExportingMetricReader({
  exporter: new ConsoleMetricExporter(),
  exportIntervalMillis: 15000,
});

const sdk = new NodeSDK({
  serviceName: "nestjs-opentelemetry-demo",
  traceExporter,
  metricReader,
  instrumentations: [getNodeAutoInstrumentations()],
});

(async () => {
  await sdk.start();
  console.log("✅ OpenTelemetry tracing & metrics initialized");
})();

process.on("SIGTERM", async () => {
  await sdk.shutdown();
  console.log(" OpenTelemetry shutdown complete");
  process.exit(0);
});

import "./tracing";
import { NestFactory } from "@nestjs/core";
import { AppModule } from "./app.module";

async function bootstrap() {
  const app = await NestFactory.create(AppModule);
  await app.listen(3000);
  console.log(" App running on http://localhost:3000");
}

bootstrap();

Adding Tracing

import { Controller, Get } from "@nestjs/common";

@Controller("orders")
export class OrderController {
  @Get()
  getOrders() {
    return [
      { id: 1, item: "Product A", qty: 2 },
      { id: 2, item: "Product B", qty: 1 },
    ];
  }
}

import { Controller, Get } from "@nestjs/common";
import { trace, context } from "@opentelemetry/api";

@Controller("products")
export class ProductController {
  @Get()
  async getProducts() {
    const tracer = trace.getTracer("nestjs-opentelemetry-demo");
    const span = tracer.startSpan(
      "fetch_products",
      undefined,
      context.active()
    );

    // Simulate a delay to mimic a real database or API call
    await new Promise((resolve) =>
      setTimeout(resolve, 100 + Math.random() * 400)
    );

    span.end();

    return [
      { id: 1, name: "Product A" },
      { id: 2, name: "Product B" },
    ];
  }
}

import { Module } from "@nestjs/common";
import { AppController } from "./app.controller";
import { AppService } from "./app.service";
import { ProductController } from "./product.controller";
import { OrderController } from "./orders.controler";

@Module({
  imports: [],
  controllers: [AppController, ProductController, OrderController],
  providers: [AppService],
})
export class AppModule {}

Adding Metrics

import {
  CallHandler,
  ExecutionContext,
  Injectable,
  NestInterceptor,
  Logger,
} from "@nestjs/common";
import { Observable, tap } from "rxjs";
import { metrics } from "@opentelemetry/api";

const meter = metrics.getMeter("nestjs-meter");
const requestCount = meter.createCounter("http_requests_total", {
  description: "Count of all HTTP requests",
});

@Injectable()
export class ObservabilityInterceptor implements NestInterceptor {
  private readonly logger = new Logger("HTTP");

  intercept(context: ExecutionContext, next: CallHandler): Observable<any> {
    const req = context.switchToHttp().getRequest();
    const method = req.method;
    const route = req.url;
    const now = Date.now();

    return next.handle().pipe(
      tap(() => {
        const duration = Date.now() - now;
        this.logger.log(`${method} ${route} - ${duration}ms`);

        requestCount.add(1, {
          method,
          route,
        });
      })
    );
  }
}

import "./tracing";
import { NestFactory } from "@nestjs/core";
import { AppModule } from "./app.module";
import { ObservabilityInterceptor } from "./observability.interceptor";

async function bootstrap() {
  const app = await NestFactory.create(AppModule);
  app.useGlobalInterceptors(new ObservabilityInterceptor()); // Register the interceptor globally

  await app.listen(3000);
  console.log(" App running on http://localhost:3000");
}

bootstrap();

npm run start:dev

Viewing Traces in Jaeger

curl http://localhost:3000/orders

curl http://localhost:3000/products

What’s Next?

Conclusion

Context Is Complex

What Are C4 Diagrams?

• Is that user insight visible in the C4 diagrams?
• Are there diagrams available?
• If not, can you help create or update them?

The Time When Draw.io Beat Figma

Influencing Impact

How to Get Started

Closing: Think in Layers

1. Installation of dependencies
2. Connecting to the OpenAI GPT model
3. Creating your first chat app
4. Streaming the output
5. Working with UserMessage, SystemMessage of LangChain
6. Working with JSON data and a model

Installation

openai>=1.0.0
langchain>=0.1.0
langchain-openai>=0.1.0
python-dotenv>=1.0.0 
langgraph>=0.1.0

pip3 install -r requirements.txt

pip3 list

Setting Up the Environment

OPENAI_API_KEY= "openaikey"
LANGSMITH_TRACING="true"
LANGSMITH_API_KEY="langsmith key"

Working with GPT Model

import os
from dotenv import load_dotenv
from openai import OpenAI
from langchain_openai import ChatOpenAI
from langchain.schema import HumanMessage
from langchain_core.messages import AIMessage

load_dotenv()

model = ChatOpenAI(
            model="gpt-3.5-turbo",
            api_key=os.getenv("OPENAI_API_KEY")
        )

   messages = [
        HumanMessage(content="Hi! I'm DJ"),
        AIMessage(content="Hello DJ! How can I assist you today?"),
        HumanMessage(content="What is capital of India?"),
    ]

• Messages sent by the user
• Contains the user’s input, questions or statements
• This message is sent to the AI model

• Messages sent by the AI model
• Contains the model’s responses, answers or generated content
• This message is included to represent the model’s response in the conversation history

        response = model.invoke(messages)
        return response

import os
from dotenv import load_dotenv
from openai import OpenAI
from langchain_openai import ChatOpenAI
from langchain.schema import HumanMessage
from langchain_core.messages import AIMessage

load_dotenv()

model = ChatOpenAI(
            model="gpt-3.5-turbo",
            api_key=os.getenv("OPENAI_API_KEY")
        )

def chat():
    try:
        messages = [
        HumanMessage(content="Hi! I'm DJ"),
        AIMessage(content="Hello DJ! How can I assist you today?"),
        HumanMessage(content="What is capital of India?"),
    ]
        response = model.invoke(messages)
        return response
    except Exception as e:
        print(f"Error: {e}")
        return None

if __name__ == "__main__":
    dochat = chat()
    print(dochat.content)

python3 main.py

Streaming Response

response = model.stream(messages)
        print("Streaming response:")
        for chunk in response:
            print(chunk.content, end="", flush=True)
            print()  

while True:
            # Ask user for input
            user_input = input("You: ").strip()
            
            # Check if user wants to exit
            if user_input.lower() in ['exit']:
                print("Goodbye!")
                break
            
            if not user_input:
                continue
            
            messages.append(HumanMessage(content=user_input))
            response = model.invoke(messages)
            messages.append(response)
            print(f"AI: {response.content}")
            print()

import os
from dotenv import load_dotenv
from openai import OpenAI
from langchain_openai import ChatOpenAI
from langchain.schema import HumanMessage
from langchain_core.messages import AIMessage

load_dotenv()

model = ChatOpenAI(
    model="gpt-3.5-turbo",
    api_key=os.getenv("OPENAI_API_KEY")
)

def chat():
    try:
        messages = [
            HumanMessage(content="Hi! I'm DJ"),
            AIMessage(content="Hello DJ! How can I assist you today?"),
        ]
        
        print("Chat started! Type 'quit' to exit.")
        print("-" * 30)
        
        while True:
            # Ask user for input
            user_input = input("You: ").strip()
            
            # Check if user wants to exit
            if user_input.lower() in ['exit']:
                print("Goodbye!")
                break
            
            if not user_input:
                continue
            
            messages.append(HumanMessage(content=user_input))
            response = model.invoke(messages)
            messages.append(response)
            print(f"AI: {response.content}")
            print()
            
    except Exception as e:
        print(f"Error: {e}")
        return None

if __name__ == "__main__":
    chat()

Working with JSON Data

data = {
    "company": {
        "name": "NomadCoder AI",
        "founded": 2024,
        "employees": 5,
        "location": "San Francisco",
        "industry": "Technology"
    },
    "training": [
        {"name": "Angular ", "price": 100},
        {"name": "React", "price": 200},
         {"name": "Lanchain ", "price": 500}
    ]
}

data_string = json.dumps(data, indent=2)

system_prompt = f"""Answer questions based only on the data provided below.
If the answer is not found in the data, reply: 'Not available in provided data.'
Data:{data_string}"""

  messages = [
            SystemMessage(content=system_prompt),
            HumanMessage(content="Hi! I'm DJ"),
            AIMessage(content="Hello DJ! How can I assist you today?"),
        ]

import json
import os
from dotenv import load_dotenv
from openai import OpenAI
from langchain_openai import ChatOpenAI
from langchain.schema import HumanMessage, SystemMessage
from langchain_core.messages import AIMessage

load_dotenv()

data = {
    "company": {
        "name": "NomadCoder AI",
        "founded": 2024,
        "employees": 5,
        "location": "San Francisco",
        "industry": "Technology"
    },
    "training": [
        {"name": "Angular ", "price": 100},
        {"name": "React", "price": 200},
         {"name": "Lanchain ", "price": 500}
    ]
}

data_string = json.dumps(data, indent=2)

system_prompt = f"""Answer questions based only on the data provided below.
If the answer is not found in the data, reply: 'Not available in provided data.'
Data:{data_string}"""

model = ChatOpenAI(
    model="gpt-3.5-turbo",
    api_key=os.getenv("OPENAI_API_KEY")
)

def chat():
    try:
        messages = [
            SystemMessage(content=system_prompt),
            HumanMessage(content="Hi! I'm DJ"),
            AIMessage(content="Hello DJ! How can I assist you today?"),
        ]
        
        print("Chat started! Type 'quit' to exit.")
        print("-" * 30)
        
        while True:
            # Ask user for input
            user_input = input("You: ").strip()
            
            # Check if user wants to exit
            if user_input.lower() in ['exit']:
                print("Goodbye!")
                break
            
            if not user_input:
                continue
            
            messages.append(HumanMessage(content=user_input))
            response = model.invoke(messages)
            messages.append(response)
            print(f"AI: {response.content}")
            print()
            
    except Exception as e:
        print(f"Error: {e}")
        return None

if __name__ == "__main__":
    chat()

Summary

• First Contentful Paint
• Largest Contentful Paint
• Total Blocking Time
• Cumulative Layout Shift
• Speed Index

Understanding Google’s Page Speed Metrics

First Contentful Paint

• Imagery
• Fonts
• SVGs
• Tables
• Forms
• Embedded content (not in iframes)
• <canvas> elements (non-white)

Largest Contentful Paint

• Time to first byte (TTFB)
• Resource load delay
• Resource load duration
• Element render delay

Total Blocking Time

Cumulative Layout Shift

• Lose track of what they’re looking at
• Click on the wrong button
• Stop reading or engaging with the page altogether, waiting to see if it’s done loading

• Asynchronous loading of resources
• Images or videos aren’t programmed with set dimensions
• Embedded content and iframes lack set dimensions
• Third-party ads dynamically resize
• Fonts render incorrectly upon initial load

Speed Index

• Render blocking requests
• Use efficient cache lifetimes
• Improve image delivery
• Legacy JavaScript
• Forced reflow
• LCP request discovery
• Network dependency tree

Wrapping Up

Getting Started

• Installing Homebrew (The Developer Toolkit)
• Setting Up Your Shell with Oh My Zsh
• Installing and Setting Up Git
• Setting Up SSH Keys
• Setting Up Node.js via NVM
• Setting Up Your First Web Project with Vite
• Setting Up Visual Studio Code
• Optional macOS Tweaks & Customizations

Installing Homebrew (The Developer Toolkit)

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

Installing Essential GUI Apps with Homebrew

# Install GUI programs
brew install --cask \
google-chrome \
iterm2 \
visual-studio-code \
rectangle

Install Terminal Tools

brew install \
wget \
git \
nvm \
pnpm \
cmatrix

Setting Up Your Shell with Oh My Zsh

sh -c "$(curl -fsSL https://raw.githubusercontent.com/ohmyzsh/ohmyzsh/master/tools/install.sh)"

omz update

Custom Shortcuts

alias x86="env /usr/bin/arch -x86_64 /bin/zsh --login"
alias arm="env /usr/bin/arch -arm64 /bin/zsh --login"

uname -m

Recommended Oh My Zsh Plugins

• zsh-autosuggestions – Suggests commands as you type, based on history
• zsh-completions – Adds extended tab-completion support
• zsh-syntax-highlighting – Highlights commands as you type, making it easier to spot errors

Installing and Setting Up Git

git config --global user.name "Your Name"
git config --global user.email "you@your-domain.com"

Add a Better Git Log View

git config --global alias.lg "log --color --graph --pretty=format:'%Cred%h%Creset -%C(yellow)%d%Creset %s %Cgreen(%cr) %C(bold blue)<%an>%Creset' --abbrev-commit"

git lg

Set main as Default Branch

git config --global init.defaultBranch main

View Your Git Configuration

git config --list

alias gitconfig="git config --list"

Setting Up SSH Keys

1. Using a single key for all services
2. Generating a separate key for each service (recommended for better security and separation)

Step 1: Create a New SSH Key

# create a new directory in case the folder doesn’t already exist
mkdir -p ~/.ssh
cd ~/.ssh
# Optional: create an alias for quick access
# alias sshhome="cd ~/.ssh"

ssh-keygen -t ed25519 -C "github"
# Recommended file name: github
# Set a secure passphrase and store it somewhere safe

Step 2: Verify the Key

ssh-keygen -y -f github
# You'll be prompted for your passphrase

Step 3: Configure SSH to Use the Key

# in case the file doesn't exist yet
touch ~/.ssh/config

Host github.com
AddKeysToAgent yes
UseKeychain yes
IdentityFile ~/.ssh/github

Step 4: Add Key to macOS Keychain

ssh-add --apple-use-keychain ~/.ssh/github

Step 5: Add the Public Key to GitHub

# Copy the public key to your clipboard
pbcopy < ~/.ssh/github.pub
# Then paste it on https://github.com/settings/keys

brew install gh # if not already installed
gh auth login  # follow the interactive prompts
# Add your SSH key
gh ssh-key add ~/.ssh/github.pub -t "GitHub Key"

• Generated a dedicated SSH key for GitHub
• Protected it with a passphrase
• Configured automatic usage in terminal sessions
• And linked it to your GitHub account

Setting Up Node.js via NVM

echo "source $(brew --prefix nvm)/nvm.sh" >> ~/.zshrc
source ~/.zshrc
# Or use an alias like:
# zshsource

Install the Latest LTS Version

nvm install --lts

node -v  # Displays the installed Node.js version
npm -v  # Displays the installed npm version

Update npm to the Latest Version

npm install -g npm@latest

Set npm Author Default (Optional)

npm set init-author-name="your name"
npm set init-author-email="you@example.com"
npm set init-author-url="example.com"

Log in to npm (For Library Authors)

npm adduser

Managing Node.js Versions

nvm list

nvm install <version> --reinstall-packages-from=$(nvm current)
nvm use <version>
nvm alias default <version>

View Global npm Packages

npm list -g --depth=0

Setting Up Your First Web Project with Vite

Setting Up Vite

npm create vite@latest  my-first-project

> npx
> "create-vite" my-first-project
│
◆  Select a framework:
│  ● Vanilla
│  ○ Vue
│  ○ React
│  ○ Preact
│  ○ Lit
│  ○ Svelte
│  ○ Solid
│  ○ Qwik
│  ○ Angular
│  ○ Marko
│  ○ Others
└

◆  Select a variant:
│  ○ TypeScript
│  ○ TypeScript + SWC
│  ● JavaScript
│  ○ JavaScript + SWC
│  ○ React Router v7 ↗
│  ○ TanStack Router ↗
│  ○ RedwoodSDK ↗

◇  Scaffolding project **in** /Users/mac/my-first-project...
│
└  Done. Now run:
cd my-first-project
npm install
npm run dev

cd my-first-project

npm install

npm run dev

VITE v6.3.5  ready in 1227 ms

➜  Local:  http://localhost:5173/

➜  Network: use --host to expose

➜  press h + enter to show help

Setting Up Visual Studio Code

Install and Configure VS Code

1. Press Cmd + Shift + P
2. Search for and run Shell Command: Install 'code' command in PATH

code .

Optional: My Personal VS Code Setup

Move Search to Bottom Panel

My VS Code Settings JSON

My Keybindings

[
{
"key": "shift+cmd+e",
"command": "editor.action.copyLinesDownAction",
"when": "editorTextFocus && !editorReadonly"
},
{
"key": "shift+alt+down",
"command": "-editor.action.copyLinesDownAction",
"when": "editorTextFocus && !editorReadonly"
},
{
"key": "shift+cmd+d",
"command": "editor.action.deleteLines",
"when": "textInputFocus && !editorReadonly"
},
{
"key": "shift+cmd+k",
"command": "-editor.action.deleteLines",
"when": "textInputFocus && !editorReadonly"
}
]

Must-Have VS Code Extensions for Web Dev

• Web Dev Pack: A curated pack for frontend developers
• Prettier: Code formatting made effortless
• GitLens: Powerful Git visualizer inside VS Code
• Bracket Pair Colorizer: Makes nested code easier to read
• Code Spell Checker: Prevent typos and misspellings
• Github Copilot: Your AI pair programming buddy
• Eslint: Real-time linting for JavaScript/TypeScript

Optional macOS Tweaks & Customizations

System Preferences

Appearance

• Enable Dark Mode.
• Set Scroll Bars to: Always show

Dock

• Remove unused apps from the Dock
• Enable Auto-hide Dock.
• Enable Show indicators for open apps.
• Enable Show recent applications in Dock.
• Go to Battery settings and enable: Show Percentage

Security & Privacy

• Turn on FileVault to encrypt your SSD.
• Under App Permissions: Allow apps from App Store and identified developers
• In Screen Recording: Add your browser

Sharing

• “Change computer name from the terminal”

sudo scutil --set ComputerName "Your-Mac"

sudo scutil --set LocalHostName "Your-Mac"

sudo scutil --set HostName "Your-Mac"

Keyboard Settings

• In Mission Control: Disable all shortcuts
• Set Fn key to: “Do Nothing”
• Key Repeat: Fast
• Delay Until Repeat: Short
• Under Text settings:
  • Disable “Capitalize words automatically”
  • Disable “Smart quotes and dashes”
  • Disable “Add period with double‑space”
  • Use straight quotes: " and '

Trackpad Settings

• Tracking Speed: Max
• Enable Tap to Click
• Under Point & Click: Disable “Look up & data detectors”
• Under More Gestures: Disable “Notification Center” gesture

Terminal Defaults (macOS)

# Show Library folder

chflags nohidden ~/Library

# Show hidden files in Finder

defaults write com.apple.finder AppleShowAllFiles YES

# Show path bar and status bar

defaults write com.apple.finder ShowPathbar -bool true

defaults write com.apple.finder ShowStatusBar -bool true

# Disable swipe-to-navigate in Chrome

defaults write com.google.Chrome AppleEnableSwipeNavigateWithScrolls -bool false

# Prevent Preview from reopening previously opened files

defaults write com.apple.Preview ApplePersistenceIgnoreState YES

iTerm2 Setup (Optional)

• Set Iterm2 as the default terminal
• Preferences -> Appearances:
• Under Windows, check: “Hide Scrollbar”
• Under Tabs, uncheck: “Show tab bar in fullscreen”
• Profiles -> Window:
• Transparency: 10%
• Style: Full Screen
• Screen: Main Screen
• Profiles -> Advanced:
• Set Semantic History -> Open with editor to Visual Studio Code

Wrapping Up

Further Resources

• Blazor Basics: Getting Started with Blazor Development in Visual Studio 2022
• React Basics: Getting Started with React and Visual Studio Code
• Vue Basics: Getting Started with Vue.js and Visual Studio Code
• Angular Basics: Setting up Angular Development Environment
• How to install Python & Django on Mac
• Learning NestJS Part 1: Getting Started
• Installing Ruby on Rail on Mac

Design Systems as Living Languages

When Perfect Consistency Fails

The Birth of a Dialect

The Flexibility Framework

The Decision Ladder

Unity Beats Uniformity

Governance Without Gates

Start Small: Your First Dialect

Beyond the Component Library

I love how these lines separate and reunite. And the fact that I can express this idea in 3 or 4 lines of code.
For me they’re lives represented by closed paths that end where they started, spending part of the journey together, separating while we go in different directions and maybe reconnecting again in the future.
#CreativeCoding #algorithmicart #proceduralArt #OPENRNDR #Kotlin

The drawing is made by choosing 10 random points, drawing a curve through those points, then slightly scooching the points and drawing another curve. There are 40 curves, each slightly different than the last. Occasionally the next curve makes a jump, which is why they separate and reunite.
Eventually I made something similar:

Along the way I had to learn about three techniques I got from the Kotlin code: Hobby curves, Hilbert sorting, and simplex noise.
Each of these algorithms tries to do something “natural” automatically, so that we can generate art that looks nice without any manual steps.

Hobby curves

To draw swoopy curves through our random points, we use an algorithm developed by John Hobby as part of Donald Knuth’s Metafont type design system. Jake Low has a great interactive page for playing with Hobby curves, you should try it.
Here are three examples of Hobby curves through ten random points:

The curves are nice, but kind of a scribble, because we’re joining points together in the order we generated them (shown by the green lines). If you asked a person to connect random points, they wouldn’t jump back and forth across the canvas like this. They would find a nearby point to use next, producing a more natural tour of the set.
We’re generating everything automatically, so we can’t manually intervene to choose a natural order for the points. Instead we use Hilbert sorting.

Hilbert sorting

The Hilbert space-filling fractal visits every square in a 2D grid. Hilbert sorting uses a Hilbert fractal traversing the canvas, and sorts the points by when their square is visited by the fractal. This gives a tour of the points that corresponds more closely to what people expect. Points that are close together in space are likely (but not guaranteed) to be close in the ordering.
If we sort the points using Hilbert sorting, we get much nicer curves. Here are the same points as last time:

Here are pairs of the same points, unsorted and sorted side-by-side:

If you compare closely, the points in each pair are the same, but the sorted points are connected in a better order, producing nicer curves.

Simplex noise

Choosing random points would be easy to do with a random number generator, but we want the points to move in interesting graceful ways. To do that, we use simplex noise. This is a 2D function (let’s call the inputs u and v) that produces a value from -1 to 1. The important thing is the function is continuous: if you sample it at two (u,v) coordinates that are close together, the results will be close together. But it’s also random: the continuous curves you get are wavy in unpredictable ways. Think of the simplex noise function as a smooth hilly landscape.
To get an (x,y) point for our drawing, we choose a (u,v) coordinate to produce an x value and a completely different (u,v) coordinate for the y. To get the next (x,y) point, we keep the u values the same and change the v values by just a tiny bit. That makes the (x,y) points move smoothly but interestingly.
Here are the trails of four points taking 50 steps using this scheme:

If we use seven points taking five steps, and draw curves through the seven points at each step, we get examples like this:

I’ve left the points visible, and given them large steps so the lines are very widely spaced to show the motion. Taking out the points and drawing more lines with smaller steps gives us this:

With 40 lines drawn wider with some transparency, we start to see the smoky fluidity:

Jumps

In his Mastodon post, aBe commented on the separating of the lines as one of the things he liked about this. But why do they do that? If we are moving the points in small increments, why do the curves sometimes make large jumps?
The first reason is because of Hobby curves. They do a great job drawing a curve through a set of points as a person might. But a downside of the algorithm is sometimes changing a point a small amount makes the entire curve take a different route. If you play around with the interactive examples on Jake Low’s page you will see the curve can unexpectedly take a different shape.
As we inch our points along, sometimes the Hobby curve jumps.
The second reason is due to Hilbert sorting. Each of our lines is sorted independently of how the previous line was sorted. If a point’s small motion moves it into a different grid square, it can change the sorting order, which changes the Hobby curve even more.
If we sort the first line, and then keep that order of points for all the lines, the result has fewer jumps, but the Hobby curves still act unpredictably:

Colophon

This was all done with Python, using other people’s implementations of the hard parts: hobby.py, hilbertcurve, and super-simplex. My code is on GitHub (nedbat/fluidity), but it’s a mess. Think of it as a woodworking studio with half-finished pieces and wood chips strewn everywhere.
A lot of the learning and experimentation was in my Jupyter notebook. Part of the process for work like this is playing around with different values of tweakable parameters and seeds for the random numbers to get the effect you want, either artistic or pedagogical. The notebook shows some of the thumbnail galleries I used to pick the examples to show.
I went on to play with animations, which led to other learnings, but those will have to wait for another blog post.

What Data Does the Government Already Have?

Augmenting Government Data with Corporate Data

Personal Targeting Using Data

The Risks of Mass Surveillance

Opportunistic Use of Data

Encryption Isn’t a Magic Bullet—But Use It Anyway

Shifting Risks of Decentralization

Being Innocent Won’t Protect You

Effective Opposition Requires Being Online