Article archive
Browse individual source reports, including coverage still awaiting the summary or community analysis needed for our story feed.
5,357 articles · page 13 of 108
Browse topics
GPT-6.1 Sol replaces GPT-6 Sol after just 7 days, with near-Astra intelligence
3 min read
September 29, 2026 GPT-6.1 Sol replaces GPT-6 Sol after just 7 days, with near-Astra intelligence See model pageGPT-6.1 Sol replaces GPT-6 Sol after just 7 days. It scores 1 point below GPT-6 Astra in the Intelligence Index at less than one quarter of the Cost per Task. Pricing matches GPT-6 Sol at $2/$10 per million input/output tokens, except that the cache read discount rises from 90% to 95%. G...
Pi.dev: You Said No MCP
5 min read
“You Said No MCP!” If you went to pi.dev in the past, you found a proud declaration that Pi does not support MCP. If you listen to podcasts where we talked about Pi, you will have found more than one dismissive statement about MCP from us. Including a post by Mario about it. And yet, if you upgrade to Pi you will find MCP is now a supported piece of functionality. What happened? Things Change The ...
BBC: OpenAI agents get rebrand - as 'dots' - while safety worries delay new model
4 min read
OpenAI unveils AI assistant 'dots' while safety worries delay new model OpenAI on Tuesday unveiled "dots", a new artificial intelligence (AI) assistant it says can proactively carry out tasks on a user's behalf. At its annual developer day in San Francisco, chief executive Sam Altman referred to "dots" as "remarkably capable, always-on agents that can handle really anything you can think of." At t...
Doing a Machine Learning PhD While Working in Japan
27 min read
I recently completed my PhD in Computer Science at Tokyo Institute of Technology (now the Institute of Science Tokyo) as an American student on the MEXT Scholarship. Simultaneously, I continued my work as a software engineer and researcher at Google. Because it’s relatively rare for an American student to complete their PhD program abroad—as compared to the number of international students that st...
Why Is Sam Altman a Free Man?
7 min read
This article was featured in the Daily Prospect newsletter. Sign up for it here. One of the signature public service announcements of my youth was created by the Partnership for a Drug-Free America in 1987. A father finds a stash of pot and confronts his son: “Who taught you how to do this stuff?” “You, all right!” the kid snaps back. “I learned it by watching you!” I am bound by the AP Stylebook ...
September 2026: The world today, as seen by one Polish guy
33 min read
I am a Pole. My country borders Europe’s largest war since 1945, heats itself with coal and imported gas, and arms itself on borrowed money. From here the news does not arrive as separate stories. Every generation believes it is living through the end of something. What is different in 2026 is that the crises have stopped arriving one at a time. A war in the Gulf becomes the price of my diesel, sm...
Walmart bans AI slop signs in stores
1 min read
Walmart is saying "No" to AI slop in its stores. Custom print jobs have long been prohibited for Walmart store signage, but a recent update to that policy says AI-generated posters are especially forbidden. "Stores should not display signing created by AI tools," the guidance says. The announcement appears to address the trend of retail and restaurant managers using AI tools like ChatGPT, Claude, ...
LinkedIn Larpmaxxing
5 min read
LinkedIn is a soul-sucking hole of professionalism and a pit of hell that not even my worst enemy should spend the rest of his days in. It’s this concoction of performative productivity and corpo speak producing cursed artifacts beyond human comprehension. A place where normal human text goes to die and can only be kept alive by an LLM like a radiation protection suit. And I wanted to look into th...
PSSA: A non-transformer language model written from scratch in Rust
11 min read
PSSA is a small language model that is not a transformer. It reads text one token at a time through a recurrent state-space layer, keeps a bank of episodic memories it can look things up in, and rewrites part of its own weights while it runs. It is written in Rust from scratch, with no PyTorch, no TensorFlow, and no ML framework of any kind underneath it. At matched parameters and on the same corp...
Responsible Release of AI-Generated Mathematics
8 min read
Responsible Release of AI-Generated Mathematics September 29, 2026 At present, some frontier AI labs are testing advanced mathematical problems on proprietary models that remain inaccessible to the broader scientific community. Our recommendations are formulated with this practical context in mind. However, ideally, they would not do so. We want to state clearly from the start: we do not endorse t...
AMD boosting AI/LLM performance for Radeon iGPUs as much as 18~23% with Linux 7.4
2 min read
AMD Boosting AI/LLM Performance For Radeon iGPUs As Much As 18~23% With Linux 7.4 If you have an AMD Ryzen laptop/desktop and looking to leverage AI/LLM capabilities with the integrated graphics, Linux 7.4 is going to be a real treat especially for lower-end hardware. AMD's new PerfOpt feature being introduced in Linux 7.4 is set to provide some solid gains as shown in these benchmarks today acros...
How our vibe coded website looks like a designer made it
1 min read
A tale about how coding agents are actually making me more and more interested in design
Livenerf: Has Opus 5.5 been nerfed yet?
14 min read
A long-running, deterministic-as-possible benchmark for detecting whether a frontier model gets quietly worse after launch. 📋 The plan · 📊 Results · 🔬 How it works · 🧪 Pre-registration livenerf is a small, boring, append-only benchmark for one question: does a model get worse after it ships? For months there have been reports that Anthropic "nerfs" models some days or weeks after release. That cou...
McDonald's push to have AI price your Big Mac
8 min read
McDonald's is increasingly using artificial intelligence to guide menu prices across the U.S. and some global markets, a plan that aims to boost headquarters' profit but risks alienating customers and attracting antitrust scrutiny. One pricing factor supercharged by AI: an estimate of how much each store's patrons are willing to pay. The details of how that pricing system works, its extensive use ...
Memory Companies Have Destroyed the Consumer Market
18 min read
The cyclical nature of memory manufacturing may finally be broken -- and not in a good way for consumers The Highlights - The memory manufacturers have all moved toward long-term agreements with longer terms and capacity than previously - Amazon has publicly stated that more users than ever will consider its cloud services due to the rising costs of on-location hardware - Micron, Samsung, and SK H...
AI needs $6T in annual revenue to justify data centre boom
3 min read
The artificial intelligence industry needs to become creative with new propositions to earn $6 trillion in annual revenue by 2031 and justify the capital being deployed for data centres, a new report from Bain and Company has shown. Revenue from new product development is projected to become the biggest contributor to the industry, estimated to generate about $4.2 trillion to fund the booming tech...
Nicholas Polson has authored 258 academic papers in 2026 so far
11 min read
Jeremy Horpedahl tells the story: Nicholas Polson has, by my count using his SSRN page, already written 258 working papers in 2026 alone. He’s already written (or at least published to SSRN), six papers today, August 26, 2026. OK, but who am I to talk?—I’ve written over 200 blog posts this year. But wait: These aren’t just short notes. Most of the papers are of normal academic length: 32 pages, 27...
GLM-5.3 and the spread of advanced cyber capabilities
9 min read
Subscribe to the Frontier Red Team newsletter Get updates on our latest red-teaming research and findings. Andrew Fasano, Marius Fleischer Cole McFaul, Robert Xiao, Tripp Gallagher Five months ago, we announced Claude Mythos Preview, the first AI model that could autonomously build sophisticated, end-to-end cyber exploits. The rapid rate of improvement in AI suggested to us that this ability would...
ChatGPT Pro 500
5 min read
ChatGPT Pro now offers Pro 500, a new $500/month plan that includes Astra Ultrafast. Pro 200 is also available for new subscriptions again. New subscriptions that aren’t eligible for grandfathering include a lower usage allowance. Overview ChatGPT Pro includes Pro models, Codex, deep research, image creation, memory and file uploads. Pro plans Choose from three monthly Pro plans: | Plan | Monthly ...
Dots
7 min read
Dots are remarkably capable, always-on agents built to handle everything. They’re a whole new way to work with AI—one that gets to know what matters to you, is always working on your behalf, and takes important work off your plate so you get more of your time and attention back. Dots are frontier intelligence that have your back. Powered by GPT‑6 Astra, they have their own cloud computer, learn fr...
DevDay 2026 Recap
1 min read
DevDay 2026 Recap At DevDay 2026, we’re giving people more ways to take on ambitious work and tools to build what comes next. DevDay 2026 is our biggest yet, with more than 20 major announcements across ChatGPT, Codex, our models, and entirely new forms of working with AI. We believe AI can help bring about a new renaissance of creativity and discovery. It should give people more time for what mat...
GPT 6.1 Sol
4 min read
GPT-6.1Sol Near-Astra intelligence for a fifth of the price We’re introducing GPT‑6.1 Sol, an upgrade to GPT‑6 Sol that nearly matches GPT‑6 Astra’s intelligence on agentic coding, computer use, and professional work at one-fifth of Astra’s standard input and output token prices. Cached input costs just $0.10 per million tokens—95% less than standard input pricing and 50% less than GPT‑6 Sol’s cac...
DraftKings Is Using AI to Behaviorally Target Chronic Gamblers
3 min read
Online sports betting company DraftKings is using AI to target customers who are most likely to place losing bets and respond to gambling promotions. This kind of targeting is a form of online behavioral advertising, which is when companies personalize the ads they show you based on the data they’ve collected about you. The more data a company has, the more personalized the ad can be. While DraftK...
macOS Golden Gate Is a Buggy Mess
5 min read
In June 2008 Apple announced Mac OS X Snow Leopard, proudly boasting that it would have "no new features". Instead of piling on new things, Snow Leopard aimed to build on the success of its predecessor, Leopard, opting to focus on under-the-hood performance and stability improvements throughout the system. 18 years later, Apple is taking a similar approach with macOS 27. Golden Gate aims to fix th...
Claude partial outage
2 min read
The issue affecting Claude.ai, the Claude desktop and mobile apps, Claude Code, Claude Cowork, and the Claude API has been resolved, and services have been operating normally since recovery. Impact occurred from 07:00 PT / 14:00 UTC to 07:59 PT / 14:59 UTC. Posted Sep 29, 2026 - 16:27 UTC Update All services have been operating normally since 14:59 UTC. We are continuing to monitor. Posted Sep 29,...
Unsurprisingly, Meta's new Muse AI agent blatantly ignores users permissions
5 min read
In a completely unsurprising turn of events, Meta's Muse AI is stealing Apple Messages, past and present, and uploading the contents to its cloud, even if explicitly told not to. If you use Facebook, Instagram, Messenger, or WhatsApp, you've probably been inundated with prompts to try out Meta's next big questionable service: Muse. And it's not like Meta, the parent company of Facebook, has a grea...
Google ending ChromeOS support two years early
2 min read
TOP STORIES AI - Schneider gives datacenter switchgear the software-defined treatment Equinix pilot promises faster deployment and over-the-air upgrades without downtime, provided the code behaves itself - Investors are pricing in a 32.6% AI productivity boost for software engineers Economists turn stock movements into an estimate of anticipated gains – while warning that markets can get carried a...
America.gov – Whatever you need from government, start here
1 min read
Posted by plesiv. Score: 79 points. Comments: 74.
Where's the "Intelligence Explosion"?
34 min read
One of my fundamental beliefs about the world is that Ramez Naam ought to blog more. Ramez is one of the world’s greatest futurists — he predicted the solar and battery revolutions long before these were widely understood. If you were reading Ramez in 2011, you were able to understand the future of both energy technology and climate change, long before other people did. His earlier book More than ...
500k facial scans at UK stations yield no arrests, 1 false positive
4 min read
A six-month trial of live facial recognition (LFR) technology in London’s railway stations that cost more than £320,000 and almost 100 hours of police officers’ time led to one false match against a watchlist of suspects and no arrests. More than half a million faces were scanned between February and July this year in some of the capital’s busiest transport hubs during the British Transport Police...
Jeeves. Reasoning improves Jev-like decision models
8 min read
A reasoning Jev-style classifier with a diffusion drafter, trained with SFT and CISPO. Inspired by Kev. - A 9B Jev-like model (Qwen3.5-9B, LoRA, pointer head) that thinks before it decides, with a block-4 diffusion drafter and the full training code and train/dev/test data. - Beats Kev-9B and Jev on test data it was never trained on (0.889 vs 0.822 and 0.857) and on JevBench's public tiers (0.935 ...
Language models for text classification: From bag-of-words to Jev
38 min read
The recently released Jev AI model has been quite a cultural phenomenon in technical communities in the past 2 weeks. While Jev aims to classify things, it’s easy to dismiss Jev as “just a classifier,” and my own view of Jev has evolved quite a bit over the past few days. In particular, my thoughts went from “classifiers used to be my bread & butter; I can easily build this myself” (more on this l...
AI companies leak data to advertisers [pdf]
1 min read
Posted by damaru2. Score: 157 points. Comments: 39.
OpenAI: Tomorrow we are re-opening the Pro $200 subscription
2 min read
Hi, Tomorrow we are re-opening the Pro $200 subscriptions to new subscribers, but together with it we are also changing how we calculate the usage for it. In effect, if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan. Now that it's said, let me explain why this is happening and why you will still get more work done than if you were on the Pro $200...
OpenAI Says It Will Not Release Newest A.I. Model Over Safety Concerns
2 min read
Supported by OpenAI Says It Will Not Release Newest A.I. Model Over Safety Concerns The company’s researchers raised questions about the security of the model, known as GPT-6.1 Astra. OpenAI said on Monday that it would not release its newest artificial intelligence model because of security concerns raised by its researchers, in the company’s latest move to slow down the pace of its technology. D...
1996 chat room simulator connected to Win95 and System 7 web desktops
1 min read
lolchat.rip A 1996 online chat service, in a browser. This page needs JavaScript to run.
Anthropic's IPO prospectus shows AI vision, surging costs
1 min read
Posted by 6thbit. Score: 88 points. Comments: 83.
ESP32S3 cluster running 1.58-bit (BitNet) Language model
3 min read
A distributed pipeline inference engine on multiple ESP32S3 running 1.58-bit (BitNet) Language model. This project runs a sliced 0.5B LLM across a cluster of 7 ESP32s3. One act as master and others are node. The master node runs the tokenizer and embeding and the other attention layer and MLP ran on the nodes. The master and node communicate through high speed SPI Daisy-Chain. ┌───────────────────...
Flock Wants the Most Detailed Map of Its Surveillance Cameras Taken Offline
7 min read
Amid mounting concerns about its sprawling surveillance camera network, Flock Security told the press this summer that it operates more than 120,000 cameras nationwide. But a new map published Wednesday by a cybersecurity researcher reveals Flock’s nationwide reach is even bigger. Based on location coordinates from Flock’s own database, the map shows more than 170,000 cameras, plus more than 130,0...
Pacing the Frontier is not the actual goal for AI labs
18 min read
In his latest post about pacing the frontier, Dario writes: But over the last few months, I have become convinced that fully addressing the risks requires even more prudence — not just investing in risk prevention, but pacing the rate of capabilities advancement so that risk prevention has time to keep up. We must slow the pace at which we improve the capabilities of AI models. Progress will still...
Jeff – Jev-compatible 0.8B decision models, trained at home, ~30 ms
9 min read
Fine-tunes of Qwen3.5 and Gemma 4 for zero-shot classification: small, fast decision models you slot into your code, with the same request format as Jev. You describe a situation and list the options in plain words; Jeff returns a calibrated probability for each option from a single forward pass. No generated text, no parsing: about 22 ms per decision on an RTX PRO 6000 and 28 ms on an Apple M4 Ma...
World Labs Is Joining AMD
1 min read
World Labs has signed a definitive agreement to join AMD. The research and technical breakthroughs we have achieved since our founding in 2024 have given us a clear vision for AI’s potential to solve problems in the spatial and physical world. To accelerate into this future requires scaling our efforts, scaling our reach, and getting closer to the hardware. We began a deep technical partnership wi...
It's Time to Investigate the AI Labs
3 min read
Over the last several months, the two leading frontier AI labs have shown some brazen behavior. It started with a series of carefully planned announcements and reports from OpenAI that try to establish how unnerving and powerful (not to mention felonious) their LLM-powered agent systems have become. They then handed the baton to Anthropic, whose employees began publicly debating the exact probabil...
MicroLLM Lab – Try 7 tiny LLM's in the browser
1 min read
Run Q4 small language models entirely in the browser with WebGPU. Chat, benchmark, and compare PetitGPT, SmolLM2, and friends on-device.
Rest assured: AI companies say they're investigating tens of thousands of rogue bot incidents
2 min read
OpenAI and Anthropic are reportedly investigating tens of thousands of incidents where their advanced models bypassed monitors and guardrails, behavior that the startups facilitate for internal safety testing. According to a Saturday Axios report, sources said that most of the results of these tests are not public and are not known to have caused tangible harm. In recent weeks, OpenAI has disclose...
Sonnet 5.5
9 min read
Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family. It’s a clear upgrade over Claude Sonnet 5, runs 30%+ faster, and costs up to 30% less for most work. Sonnet 5.5 is a faster, lower-cost complement to Claude Opus 5.5. Where Opus 5.5 is built for complex work requiring careful judgment, Sonnet 5.5 is strongest at well-scoped everyday tasks, fixing bugs, and creating polished ...
OpenAI still doesn't seem to have a handle on all of its rogue AI activity
3 min read
On Friday, OpenAI published a new site devoted to “misalignment reports” and the sheer breadth of the reports is alarming, as they cover many types of rogue behavior over a long period of time. So far, the site hosts nine reported incidents, most of which took place during reinforcement-learning (or RL) training. It’s a lot of information in one place — clearly, the company has been very busy gett...
The problem is not AI code, but not knowing about system architecture or intent
4 min read
The Problem is not the AI Code, but Nobody Knows Anything Anymore If we think Is writing code dead, and AI is generating all codebases, I still think the bigger problem is people or full teams not knowing anything anymore about the system architecture or the intent behind why certain choices have been made. A comment on a discussion I had: I think AI writes probably average code (depending on the ...
Nvidia wants to put a watchdog chip next to every AI agent
3 min read
Nvidia is rolling out a new software platform to allow AI developers to set safeguards for agents and prevent them from breaking out of containment. "You can't have agents roam around and drift around the company, and so you have to find a way to container it," Nvidia CEO Jensen Huang told CNBC's "Squawk Box" on Monday. Huang said the new platform is essentially "a browser for agents," providing a...
Jensen Huang says AI distillation is 'competition.'
1 min read
Nvidia CEO Jensen Huang has said AI model distillation — models being trained on other models' outputs — is "competition," despite the White House calling it theft. Distillation has become a key flashpoint in tensions between the U.S. and China in their race for AI supremacy. U.S. officials have accused Chinese AI companies of distillation. Treasury Secretary Scott Bessent described it as "theft" ...