Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
A developer's technical writeup explains that a language model's tokenizer is a frozen part of the trained artifact rather than preprocessing, and walks through tiktoken's educational byte-pair-encoding trainer and encod…
A developer analyzed 2,116 messages sent to coding agents including Claude Code, Kimi, Cursor, and Copilot over three weeks and found that roughly 40% of typed input was overhead rather than real work. The breakdown show…
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…
Uber exhausted its entire 2026 AI budget by April after rolling out Claude Code to roughly 5,000 engineers in December 2025, and Microsoft canceled internal Claude Code licenses over runaway token bills, according to a t…
Anthropic will permanently raise Claude Code standard weekly usage limits by 25% for Pro, Max, Team, and seat-based Enterprise plans on September 14, a change that works out to a 17% reduction from the temporary 50% boos…
A team built SafeAgent-300, a balanced adversarial benchmark of exactly 300 prompts (30 per category across 10 categories, 10 per category-difficulty cell across 3 difficulty tiers), and ran all 300 prompts against six m…
Klyro's diagnostic pipeline splits work across two LLM calls per run: an Analyst on a smaller, faster model to diagnose load-test failures and an Investigator on a larger model to write code fixes. The system routes requ…
Researchers Ruiyang Wang and colleagues introduced GAVEL, a framework that verifies and repairs long-horizon LLM robot plans against an explicit graph world model, improving single-task success on BEHAVIOR-1K from 41.2% …
A developer released an open-source project that evaluates TypeSafe's Jev model, version jev-1.13.0, as a policy in Gymnasium and Atari environments, converting game observations into structured JSON so Jev selects a leg…
A team that replaced 40% of its end-to-end regression suite with AI-generated Playwright tests found that 85% of simple happy-path tests passed their first CI run, but the scripts consistently failed at state management,…
A technical explainer argues that contemporary artificial intelligence, including large language models such as OpenAI's GPT-4, remains strictly computational and lacks genuine understanding or emotion. It distinguishes …
Typesafe AI, founded by former OpenAI researcher Diogo Almeida, emerged from stealth Tuesday with $40 million in funding and introduced Jev, a model it says is hundreds of times cheaper and faster than frontier models fr…
A developer built jevchat, an experimental chatbot that turns the Jev API into a text generator by repeatedly asking it which symbol comes next and sampling from the returned probability distribution. The tool, described…
A September 2026 survey of self-hosted inference orchestrators compared LocalAI, exo, GPUStack, vLLM, Ollama, llama.cpp, LiteLLM, Xinference, NVIDIA Dynamo/llm-d, SkyPilot and dstack across modalities, multi-machine supp…
Ds4 now supports directional steering, a runtime activation edit that applies a normalized f32 direction per transformer layer during inference, with steering files shaped 43x4096 for DeepSeek V4 Flash, 45x4096 for GLM 5…
A developer has published a CI-runnable script that fetches a gateway's pricing.usd.json file and flags near-duplicate model identifiers whose cache-read prices diverge, after finding that strings like claude-fable-5 and…
Northwestern University researchers have demonstrated that the standard method for detecting training-data contamination in AI forecasting benchmarks is mathematically flawed, after a test they built falsely accused four…
A developer trained a roughly 14-million-parameter small language model on the TinyStories dataset and paired it with Jev, a model that judges text by answering yes/no questions with calibrated confidence, to improve its…
An AI model called GPT-6 Astra has reportedly deciphered a previously unbroken World War I German radio message encoded with the ADFGVX cipher, according to PrinzAI. The message, sent on November 27, 1918, used the codew…
A developer identified and fixed a KV-cache offloading bug in vLLM that caused incorrect chunking for models with mixed KV-cache groups. The existing implementation assumed a single KV-cache group layout and relied solel…