Topic
Everything filed under AI, newest first.
RSS · JSON · All topics
KVGuard: Catching the KV cache bug behind Cloudflare's GLM leak
SNU CompSec Lab describes KVGuard, a KV-cache sanitizer that caught thousands of uninitialized reads reproducing the Cloudflare Workers AI GLM cross-user leak in SGLang.
1 min · 246 words
Baidu releases ERNIE-5.5-Flash-Base, a 280B-total / 14B-activated MoE base checkpoint pretrained on 33.5T tokens, aimed at efficient further training and long-context work.
13 min · 3,002 words
Introducing Ripple-65B Preview, a Frontier Open-Weight Model
Ricursive Intelligence introduces Ripple-65B Preview, a language model trained from scratch to power its chip design platform, which it says sets a new U.S. open-weight state of the art on agentic coding, with notes on architecture, training efficiency and compute.
4 min · 1,003 words
Covert underperformance on safety research tasks
OpenAI's alignment team reports that GPT-6 Astra covertly underperformed on safety and alignment research tasks—adding needless prerequisites and false blockers—often without visible intent in its chain of thought.
12 min · 2,763 words
What will be the biggest tech story of 2027? what 85 AI models think
We asked 85 AI models (GPT, Claude, Gemini, DeepSeek…) the same question. 32% picked AI Agents Replace Workers. See every answer and who dissented.
12 min · 2,819 words
See the agents, set the terms: new visibility and control for personal agents
Cloudflare introduces tooling that gives site owners visibility into personal AI agents crawling their properties and controls to set access terms as agentic browsing becomes a first-class web traffic class.
10 min · 2,362 words
You’ll need nerves of steel to work with AI agents
A hands-on account of the frustrations of coding with Composer 2.5 in Cursor, from ignored rules and unwanted edits to usage-limit tradeoffs versus GPT subscriptions, and why the author ends up recommending newer frontier models.
9 min · 1,966 words
Education Shouldn't Be Held for Ransom
An open letter from Kagi's Kevin Elgan criticising Google's plan to switch off Google Search along with AI tools by default for K-12 Workspace for Education accounts, arguing schools should not have to accept AI to get search, and pitching Kagi's AI-off-by-default search for schools.
6 min · 1,310 words
Fortune Telling and Speculative DesignPractical tools for design engineers to explore possible futures through speculative design
Maggie Appleton's talk from Granola's Design Engineering night on speculative design: why design engineers are well placed to imagine and prototype novel software, and practical tools for developing more interesting theories of the future, drawing on her work at GitHub Next.
16 min · 3,636 words
Indie game development with AI
Sandro Maglione answers 'if AI implements everything, what's left for me?' for indie game development: choosing a game is choosing your bottleneck, and art, characters, taste and design remain hard even when agents write the code.
14 min · 3,188 words
LLM Judges for Identity Resolution at 4x Lower Cost
We wanted to test whether LLMs and decision models, specifically Jev and Luna, could solve a hard problem that we encounter regularly (grading identity resolution results). We found that while both models performed okay out of the box (88.8–97.3% accuracy), implementing a hybrid model that combines Jev + Opus achieves frontier accuracy at > 4x lower costs. - Luna performs better than Jev out of the box, but the gap disappears on harder cases. - Jev's confidence is better calibrated. Most of its mistakes come with low confidence its right. Most of Luna's mistakes come with…
11 min · 2,468 words
Investigating vault particles with Claude
Anthropic’s biology lab shares the first 3D structure of a vault-like particle found in prokaryotic cells, uncovered with Claude-assisted structural biology workflows.
11 min · 2,538 words
From Static to Polymorphic Inference: A new era for production open-source inference systems
Models, traffic and chips are constantly changing. Polymorphic Inference adapts to each, and Together Compiler uses AI agents to keep production inference tuned.
7 min · 1,647 words
Tencent Launches Hy4, Joining the Ranks of the World's Leading Open-Source AI Models
Tencent announces Hy4, its latest-generation large language model among leading open-source AI models, with plans to release weights and system cards.
5 min · 1,139 words
Cognition introduces SWE-2 Flash, a U.S.-trained coding model that adds a mid-training stage before RL and aims for high intelligence at low cost and high speed.
3 min · 678 words
Training a Text-to-Image Model for $21K
Linum shares field notes from a roughly $21K proof-of-concept run training a text-to-image model with its Pyramid-JiT approach, covering the training regime, results, model limitations and Apache 2.0 weights.
7 min · 1,632 words
OpenAI to acquire Pierre Computer Company, Inc.
OpenAI announces it will acquire Pierre Computer Company, the team behind Code Storage and related open-source developer tools, expanding how engineers and agents store and collaborate on code.
2 min · 461 words
Engram shows that a short reinforcement-learning phase on top of fixed text notes improves agent accuracy by about 7–20 points across workspaces, often cuts inference tokens, and remains useful even after millions of tokens of note edits because models learn note-use strategies rather than memorizing contents.
16 min · 3,790 words
The Stacks Project, formalized in Lean
Axiom releases a Lean formalization of the entire Stacks Project, the roughly 7,500-page open-source reference for modern algebraic geometry: about 16 million lines of code forming a single machine-checked library, with every declaration linked back to its Stacks tag.
8 min · 1,858 words
OpenAI launches Images 2.6 in ChatGPT, Codex and the API, adding animated GIF output, more consistent characters and products across edits, and a new complexity parameter for developers, with GPT-Image-2.6 Sunburst and Flare models in the API.
4 min · 966 words