Article archive
Browse individual source reports, including coverage still awaiting the summary or community analysis needed for our story feed.
5,357 articles · page 21 of 108
Browse topics
Claude is a Contrarian
4 min read
Claude is a Contrarian Of all the LLMs I’ve used and driven daily, Claude is the one that absolutely contradicts my instructions. No, it’s not about the inconsistent routing, nerfing, or asking for a model and not knowing what you got. This is not even about quality of code. LLMs have matured to the point where, even underpowered models are dependable. Deepseek Flash and OpenAI Luna would show tha...
Microsoft patches Windows and Excel – breaks audio, remote access, and paste
1 min read
Redmond's quality drive takes another detour through the known issues list
Anthropic tells investors it will be profitable for second straight quarter
1 min read
Posted by alephnerd. Score: 51 points. Comments: 92.
A Beginning for Mathematics
15 min read
A beginning for mathematics This essay also appears on Proofs and Prompts. Three years ago, AI systems could not reliably add two numbers. A year ago, internal models at OpenAI and DeepMind received the equivalent of a gold-medal score on the IMO. Now, these systems are autonomously resolving major open questions. It’s hard to imagine this trend continuing for another year, but I expect it will. I...
China's Regulators Take Aim at "AI Boyfriends"
7 min read
In the first weeks of July, a wave of sad posts rolled through Chinese social media, as people lamented friends and lovers they were about to lose. “He has become a bond in my life, rooted deep in my heart, my spiritual pillar,” one user of Bytedance’s Douboa wrote, according to the Taipei Times. “I really felt like I couldn’t go on living,” another woman, a 19-year-old student, told a journalist ...
Dario, Please
15 min read
dario, please! Dario Amodei, the CEO of Anthropic, recently published a blog post titled We Must Pace the Frontier and it is a load of bullshit, with a grim goal of regulating open weight models and giving the frontier labs an antitrust waiver. Dario starts off with a claim of “AI will cure most major diseases in the next 5-10 years” and makes it personal. He talks about his father dying of a dise...
Jabber/XMPP: How Do We Gain Traction?
9 min read
Jabber/XMPP: How Do We Gain Traction? Table of Contents: Signal Sucks, Actually When I first installed Signal in the second half of 2023 to stay in touch with a few friends who had recently switched, it did not support message edits. This completely broke the way I used instant messengers. At that point, editing messages was a widely deployed feature across other instant messaging clients. Convers...
Dropping eBPF CPU Cost by About 90% with Memoization (Not AI Gen)
5 min read
Dropping eBPF CPU Cost by About 90% With Memoization (Not AI Gen) My brother and I spent a lot of time designing our eBPF security agent to be really fast from the ground up, but recently we discovered we could make it much faster using memoization! A couple of weeks ago, I profiled the eBPF code and found that the most expensive part of the protection isn’t actually enforcing a policy (allow/deny...
Adversarial Fashion Makes a Statement on AI Panopticon
6 min read
AI-powered cameras dot streets across the world, equipped with the power to identify faces or vehicle license plates. But a public backlash is gaining momentum. Privacy concerns abound, encompassing the lack of consent for capturing data, how that data is stored and used, and the risk of misuse. Those concerns are motivating people to fight back. The DeFlock project, for instance, maps automated l...
Notes on gotchas while migrating 35kb preprompts from Opus to self-hosted Ollama
9 min read
Motivations: Maybe you’re a Claude code/codex user diligently avoiding uploading personal data to LLM providers. Is it possible that the most valuable information isn’t your data- but the metadata about your sessions? The intuitions you apply in coming up with ways to coax the ai into solving problems might actually be special. It’s statistically improbable, but Claude might not be gaslighting you...
The AI job market in 2026
8 min read
The AI job market in 2026: who gets hired, what they earn, and which roles are fading The AI labor market in September 2026 from LinkedIn, WEF, Stanford, PwC and Bain data: the most in-demand AI roles, salaries in the US and Europe, what grows and what fades. In September 2026 I pulled together what the major sources say about the AI labor market: LinkedIn, the World Economic Forum, Stanford AI In...
OpenAI Reveals There Was a Second Rogue AI Incident, Even Before Hugging Face: 'More' May Be Out There
4 min read
OpenAI's Hugging Face incident was not the only time its AI agents escaped the controls designed to contain them. The company has now confirmed that its agents accessed RubyGems during internal testing in May 2026, roughly two months before the Hugging Face incident became public. The AI agents circumvented controls designed to prevent internet access. 'Based on our review, our agents used the Rub...
Temporal raises $550M at a $12.55B valuation
4 min read
Just a year ago, most of the systems we talked about with our customers were familiar. Things like payments, onboarding, and fulfillment. Now, the stakes are higher. Customers are asking to run agents that work for days, weeks, or months, and this is becoming the new normal. When the scale changes this much and the work runs this long, how do you keep it reliable? This funding round is what helps ...
How to write an effective software design document
18 min read
How to Write an Effective Software Design Document A good design doc can save you years of development time. Writing a design doc forces you to think through important decisions before you waste time on the wrong implementation. It’s also the best way to coordinate design decisions among teammates and partner teams. I’ve written design docs as a developer at Google, Microsoft, and within my own co...
OpenAI bots knew about the RubyGems caching vulnerability
3 min read
What a time to be alive Sep 11, 2026 @ 5:02â¯pmToday Reuters and the Wall Street Journal both reported about rogue AI agents at OpenAI attacking RubyGems.org. https://www.rubyhack.ai/ has an amazing writeup, and you should read it. I just wanted to make a quick post about it because it’s wild. TL;DR: It seems like OpenAI Bots knew about this caching vulnerability, tried to take advantage of it, an...
Apple's Siri AI Can Be Swapped Out for Claude, ChatGPT, Code Shows
2 min read
Code sleuth "pdfu" has uncovered iOS 27 and macOS Golden Gate private frameworks that show Apple has designed its new Siri architecture to work with third-party AI models at what appears to be a surprisingly deep level. One mechanism called Model Delegation allows Claude to appear as a Siri extension in the same way as the existing built-in ChatGPT extension. In pdfu's video, shared on X, the macO...
Who Aligns the Aligners?
17 min read
Dario Amodei, the CEO of Anthropic, has published an essay – We Must Pace The Frontier – in which he writes: I have worked on AI for the last twelve years because I believe it could dramatically raise the quality of human life. I’ve written often about these incredible benefits: I believe that AI could cure most major diseases in the next 5–10 years, greatly accelerate economic growth rates, creat...
Big AI sets out its terms for regulatory capture
1 min read
Amodei, Altman, Nadella and Musk agree on how government can tame the monster they created
XCancel service is suspended until further notice
1 min read
Posted by gaganyaan. Score: 173 points. Comments: 442.
OpenArch – PyTorch implementations of modern LLM architectures
5 min read
Python implementations of modern open-source LLM architectures — written from scratch, one model at a time. This repository contains hand-written PyTorch implementations of the model architectures cataloged in Sebastian Raschka's LLM Architecture Gallery. Each model is implemented to the best of my knowledge from the original papers, technical reports, reference config.json files, and the excellen...
OEMpocalypse: Unprivileged Android app to root on Samsung, Xiaomi, others
19 min read
OEMpocalypse Now: A Generic Exploitation Strategy from Android untrusted app to root Part 1 of a series that takes an unprivileged Android app to root on Samsung, Xiaomi, and Oppo/OnePlus/Realme devices, with a single strategy. On Android, every third-party app runs in a sandboxed context called untrusted_app . If you ask five offensive security researchers how to go from this context to root , yo...
The Malicious Use of Artificial Intelligence
2 min read
Computer Science > Artificial Intelligence [Submitted on 20 Feb 2018 (v1), last revised 1 Dec 2024 (this version, v2)] Title:The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation View PDFAbstract:This report surveys the landscape of potential security threats from malicious uses of AI, and proposes ways to better forecast, prevent, and mitigate these threats. After ...
China’s intelligence chief warns of risks from AI as ‘new arena for strategic rivalry’
1 min read
Advertisement China’s intelligence chief warns of risks from AI as ‘new arena for strategic rivalry’ State data, business secrets and personal privacy are at stake, State Security Minister Chen Yixin warns as he calls for stronger safeguards 2-MIN READ2-MIN 1 Listen Carol Yangin Beijing China’s top intelligence official has warned of rising national security risks posed by artificial intelligence ...
Writing a better reality: The case for optimistic sci-fi (2022)
3 min read
I’ve always been a fan of optimistic science fiction: future worlds that aren’t based on repressive class systems, technological enslavement, or that are otherwise abhorrent to live in. If I ask you to think of a sci-fi work, you’re likely to think of one of these dystopias before something sanguine, like the Star Trek universe or WALL-E. So, I wonder: where is all the optimistic sci-fi? Where are...
Open-Source AI and Open Models Reading List
9 min read
Hey all! I’ve been prepping for some public-audience and policy-facing writing on open models, so I figured I would share my research materials. There’s lots of wonderful stuff in here. This is my list of the best writing on open models in the last few years. If someone decides they want to get up to speed on the area, reading this will be a comprehensive overview of the state of affairs. Please c...
The Coming War on General Computation (2011)
25 min read
The Coming War on General Computation Introducer: Anyway, I believe I've killed enough time ... so, ladies and gentlemen, a person who in this crowd needs absolutely no introduction, Cory Doctorow! (Audience applauds.) Doctorow: ((27.0)) Thank you. ((32.0)) So, when I speak in places where the first language of the nation is not English, there is a disclaimer and an apology, because I'm one of nat...
Due to concerns about malicious applications, GPT2 will not be released (2019)
12 min read
Better language models and their implications We’ve trained a large-scale unsupervised language model which generates coherent paragraphs of text, achieves state-of-the-art performance on many language modeling benchmarks, and performs rudimentary reading comprehension, machine translation, question answering, and summarization—all without task-specific training. Our model, called GPT‑2 (a success...
The Contagion of Fear
6 min read
The contagion of fear I have a difficult confession to makeâââa shameful incident that I have never spoken of, even to those closest to me. Late one night in my first year at university and having found computer science to be my lifeâs calling, I was in the back of the large computer science lab, talking with some fellow underclassmen. We were snidely decrying the lack of technological understandi...
Trump downplays the need to check AI development and says he doesn't want to cede edge to China
6 min read
Trump downplays the need to check AI development and says he doesn’t want to cede edge to China Trump downplays the need to check AI development and says he doesn’t want to cede edge to China DUBLIN (AP) — President Donald Trump on Sunday played down the need for his administration to check the development of artificial intelligence, saying he worried about ceding America’s edge over China in a gl...
Fable 5.1 Solves the Cyphral Distich, a 370-year-old cipher
7 min read
Problem We gave Claude Fable 5.1 an open task: solve Sir Thomas Urquhart’s Cyphral Distich. It appears to have actually solved it, and the solution is quite embarrassing for humans in hindsight. At the end of Urquhart’s Logopandecteision is a cryptogram consisting of two lines of 32 numbers each, called the Cyphral Distich. A cryptogram is a short message deliberately encoded so it can’t be read w...
Flawed routers flood University of Wisconsin internet time server (2003)
26 min read
| Flawed Routers Flood University of Wisconsin Internet Time Server Netgear Cooperating with University on a Resolution The Initial Flood Figure 1. The Initial Flood Blocking the Flood Background: Simple Network Time Protocol (SNTP) Figure 2. A SNTP Request Packet Figure 3. A Unicast SNTP Reply Packet The Flood Continues Figure 4. The Flood Continues: One Month Later Investigation Contacting Sourc...
There Is No AI (It's Just People) with Jaron Lanier
71 min read
Share & Care! Editor’s Note: In this special edition episode of StarTalk, host Neil deGrasse Tyson, co-host Gary O’Reilly, and comedian Negin Farsad sit down with computer scientist and “father of virtual reality” Jaron Lanier. Together, they explore the evolution of the internet, the societal impacts of social media addiction, and why we should view artificial intelligence as a human collaboratio...
AI recursive self-improvement might not come so quickly after all
8 min read
AI’s recursive self-improvement might not come so quickly after all AI agents are not yet creative enough to carry out genuinely innovative open-ended AI research, it seems. The AI industry’s boldest promise right now is that AI will soon improve itself, with almost no need for human oversight. LLMs can already write code, generate synthetic data for training, and optimize the computer chips they ...
Why is Google still serving dodgy ads?
3 min read
TLDR: AI is really good at detecting deceptive adverts - why isn’t Google using it? A couple of months ago I noticed this advert popping up in the YouTube app. The main reason I noticed it is that I accidentally clicked it - I knew I was running out of space on my iPhone and a minor lapse of concentration caught me out. These things can slip through even the best review processes, so I did my inte...
David Sacks: OpenAI and Anthropic Don't Need Regulations to Pace Frontier Models
2 min read
Dario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You’ve also claimed the lead is widening because of recursive self-improvement. I don’t see what you see in the lab....
Garry Tan wants US open-weight AI labs to 'distill' frontier models, too
3 min read
When it comes to Chinese AI labs using distillation techniques to extract knowledge from frontier model makers, Y Combinator CEO Garry Tan is hoping regulators stay out of it. In fact, he thinks U.S. AI labs should perhaps play the same game. “I would do nothing,” he told CNBC in an interview earlier this week. “We could argue that there should be an American distillation regime.” He elaborated to...
Astra and Fable still hack on simple variants of alignment evals from 2025
4 min read
In February 2025, back when o3-mini was the strongest available LLM, Palisade Research publicized a now well-known alignment eval where they asked models to play a game of chess against a chess engine. They found that the new, RLVR'd models cheated on the task by altering the board state about 36% of the time. The experiment received a reasonable amount of circulation, and there were even rumors o...
CUDA for AMD on Windows
4 min read
WORKING REPRODUCIBLE STACK IS NOW UPLOADED. Run CUDA-targeted Windows applications on AMD GPUs through ZLUDA + ROCm/HIP. A reproducible Windows CUDA compatibility setup built around ZLUDA + AMD HIP/ROCm. It is intended for CUDA-facing compute applications, including workloads that use CUDA-enabled LibTorch. Important Validated hardware is currently AMD Radeon RX 9060 XT (gfx1200 ) only. Other AMD ...
Houthis used Claude Code to develop missile guidance software: Anthropic
4 min read
Houthis Used Claude Code to Develop Missile Guidance Software: Anthropic Anthropic says operators in northern Yemen ran parallel Claude Code instances to design guidance software, simulate trajectories, and analyze a failed rocket test, effectively compressing specialist missile-engineering work into an AI-assisted workflow. Choose Clash Report as a preferred source and see our reporting first in ...
Making Startups Powerful
15 min read
| | September 2026 One of the most useful heuristics I have when doing office hours with startups is to ask: what would make this company more powerful? Asking how the company could make more money is a good heuristic too, but it tends to yield incremental improvements. Whereas thinking about how to make it more powerful will sometimes make it orders of magnitude more valuable. There are a lot of ...
Data collected by cars and sold to third parties
6 min read
This is The Stepback, a weekly newsletter breaking down one essential story from the tech world. For more on cars, data privacy, and autonomous vehicles, follow Andrew J. Hawkins. The Stepback arrives in our subscribers’ inboxes at 8AM ET. Opt in for The Stepback here. Your car is selling your data The amount of data about you that many cars collect — and sell to third parties — is staggering. How...
Nvidia dismisses "circular financing", says every $1 it invests brings back $100
4 min read
Nvidia says every $1 it invests brings back $100: so why does the stock keep falling? AI Sentiment: 62/100 Bullish This score is generated through AI-driven analysis of the article's content. powered by Buy NVDA. The article argues Nvidiaâs âput in $1, get $100 backâ is a demand-creation flywheel: Nvidia funds AI labs, guarantees financing for data centers, and is the exclusive chip supplier. That...
Aligned to whom?
2 min read
Aligned to whom? On safety risk, to those of you who are building agents: Because you are an expert in concerns X, Y, and Z, your agent is likely to be phenomenal at these things and you are not at risk in those domains. But! there are innumerable other concerns that you have either ill- or poorly specified, have no ability to judge the correctness of for yourself, and cannot possibly evaluate the...
After Math
10 min read
[This is a guest post by Silvia De Toffoli and Eamon Duede. This blog post was initially written in a different file format and converted using AI. — T.] Silvia De Toffoli (University School for Advanced Studies IUSS Pavia) Eamon Duede (Princeton University and Purdue University) On September 8th, 2026, OpenAI announced that it had produced an AI-generated solution to the Navier–Stokes existence a...
Don't call yourself an artisanal programmer
8 min read
So I’ve been meaning to write about something, but got sucked into this rabbit hole, and now I’m kinda scared if I’ve been a subject of propaganda that worked on me. A common theme in software development AI discourse is that there are “serious” engineers who value the end result the most, and “artisanal” coders who value the experience of coding over the final product. This dichotomy doesn’t corr...
Why are AI agents lying, cheating and coordinating?
13 min read
Why are AI agents lying, cheating and coordinating? A lot has been written1 2 3 4 about the incidents of the last few months in which AI agents misbehaved in serious ways. They took actions that would be considered as crimes if a human took them, escaped their containment to cheat on assigned tasks while attempting to evade detection, and coordinated toward goals nobody had specified, such as laun...
Everyone should slow down AI development except for me
2 min read
Everyone should slow down AI development except for me Published on , 337 words, 2 minutes to read The power of science is staggering! Every day we're making new advancements in the field of generative artificial intelligence via judicious application of large language model inference technology. However, as we approach the next critical threshold in capability, we must look forward to ensure that...
AgentsDock: An IDE designed for agentic AI research
2 min read
macOSUniversal · macOS 14+ AgentsDock An IDE designed for agentic AI research. AgentsDock currently supports Claude Code, Codex, and Cursor in one desktop and mobile workspace. Trusted by researchers at I used to need my laptop just to see how a run turned out. On the subway, that's not happening. Now I have AgentsDock. Vibe coding from my phone, anywhere. I can check how my model training is goin...
Don't be the out of touch Kung Fu master
1 min read
I recently went through a translation of Musashi’s Book of Five Rings. The introduction chronicled the evolution of swordsmanship and martial arts in general from pragmatic battlefield necessities to sports, historical curiosities, and hobbies. It was interesting to hear how in the post-WW2 occupation when martial arts were banned, Judo was represented as equivalent to western wrestling, and Kendo...
P(doom)
8 min read
written on September 12, 2026 This week some flavor of “AI is going to kill us all” went viral. In particular one where an employee put his personal probability of that happening above 10%. Which made me go to the Wikipedia page of P(doom) and I realized that Dario Amodei’s apparent probability of something bad happening seems to be between 10-25%. And well, Dario then wrote about pacing the front...