Full-text search across 68271 articles. Combine with topic and date filters; results sorted by relevance.
AMD researchers introduced Zebra-HyLo, an upcycling recipe that converts pretrained Transformer models into long-context hybrid LLMs through post-training on roughly 10B tokens, without pretraining from scratch. The meth…
Shrijith Venkatramana, an engineer building the LiveReview AI code review tool, published a technical explainer on LLM quantization showing that activation outliers concentrated in a handful of feature dimensions can con…
Cognition launched command-line controls and full SSH access for Devin Cloud on Monday, letting developers start, steer, resume and directly enter the remote virtual machines used by its coding agent, the company said in…
HeyGen released Code2Video on September 18th, a benchmark of 168 human-authored motion-graphics briefs that tests whether coding agents can produce watchable motion design, scored by a HeyGen-trained judge model. HeyGen …
Rijul, a developer building the AI code review tool LiveReview, explains why retrieval-augmented generation systems fail to surface information that is present in a source document. The core issue, he writes, is chunking…
A developer built >_Fa, an open-source agent harness written in Dart, by consuming roughly 40 billion tokens across cheap models like Kimi K3 and GLM-5.3 at a cost of about $17,000–$20,000 over three months, with no huma…
SpaceXAI's Grok 4.7 scored 46 on the Artificial Analysis Intelligence Index, a 2-point gain over Grok 4.6's 44, but still trails GPT-5.6 Sol (47), Meta's Muse Spark 1.3 (48), and Anthropic's Claude Fable 5 (50). The mode…
Z.ai apologized and said it fixed a default-enabled codebase indexing feature in its AI coding tool ZCode after a developer discovered a 313MB encrypted snapshot on September 18 containing roughly 345MB of a commercial p…
StepFun released Step 5 Preview, a flagship model for agentic work that natively supports text, image, and video input and offers a 1M-token context window. StepFun says the model delivers frontier-level performance acro…
A developer has proposed a "receipt"-based approach to verifying AI coding assistant setup, arguing that onboarding screens and green checkmarks provide no evidence that a model actually modified files or ran tests. The …
Independent developer Tom (writing on his personal site) released version 2 of his "hundred poster prompts" collection, rebuilding the AI image prompts after Hacker News commenters flagged recurring flaws in version 1. T…
A developer created LSP-ember, a Language Server Protocol bridge that brings autocompletion, go-to-definition, and hover support to Ember's .gjs and .gts template-tag files in Sublime Text. The package mediates between S…
Jared Palmer, vice president of engineering at Cognition, released a Qwen3.5-based generation of the open-source Kev decision-model project on September 20th, comprising 0.8B, 4B and 9B parameter variants that return typ…
A developer's fifth installment in an enterprise AI adoption series outlines Stage 5, governance and observability, arguing that logging and budget limits have run inside the gateway since Stage 1 while this stage adds c…
Rijul, a developer building the AI code review tool LiveReview, published an explainer comparing two methods for aligning language models with human preferences: Direct Preference Optimization (DPO) and Reinforcement Lea…
In the third essay of his AI Harness Engineering series, developer Derek Wang argues that architectural decisions made only in chat with AI coding agents do not persist, since they evaporate when the context window close…
The Berkeley Overmind, an AI agent built by a UC Berkeley team led by computer science professor Dan Klein, defeated human StarCraft expert Oriol Vinyals for the first time after dozens of test matches, three days before…
The Federal Register, run by the National Archives, removed an AI search tool built on Alibaba's Qwen model on Wednesday, September 17, 2026, the same day social media posts flagged it, according to Reuters, which review…
Anthropic and Accenture announced a multi-year partnership to move Claude-based AI deployments from isolated pilots into production, centered on a new Accenture Anthropic Business Group. Accenture plans to train roughly …
Anthropic reported on September 17 that its Claude model now leads 26% of the company's AI research and development work, up from essentially zero in February, with Claude authoring more than 80% of code merged into Anth…