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 reports that Anthropic's Claude Code CLI tool cut feature development time from 4-6 hours to 1-2 hours by reading entire codebases and making coordinated multi-file changes. The engineer says the tool, which …
A fine-tuned Qwen2.5-Coder-14B-Instruct model compiled 170 of 184 MQL5 test prompts after one epoch of training on the author's dataset, up from 2 of 184 for the untouched base model, according to a published benchmark c…
A developer ran a controlled experiment with 40 agents (later replicated with another 40) to test whether the presence of a language model causes engineers to patch symptoms rather than trace root causes. Replacing the L…
A developer argues that enterprise AI governance must be runtime and deterministic, positioning their SAFi harness as an alternative to agentic frameworks like Claude Code. In SAFi, the LLM is treated as a substrate rath…
AI models behave probabilistically rather than deterministically, meaning the same input can produce different outputs on separate runs, according to an explainer on how AI models work. The explainer notes that models ge…
A technical explainer on AI agents details how autonomous loops chain tool calls through repeated cycles of deciding, acting, and reviewing results until a task completes or a stopping condition halts them. The piece ide…
Large language models hallucinate because they are trained to predict plausible next words rather than verify facts, producing false outputs such as nonexistent library methods, wrong import paths, and config flags remov…
AI benchmarks such as BioMysteryBench, GeneBench, and Terminal-Bench 2.1 are inconsequential at best and misleading at worst, according to a deep dive into the scores cited for GPT-5.6, Fable 5.1, and Opus 5. Anthropic r…
Large language models process text as tokens rather than words, and providers charge separately for input and output tokens, usually in price-per-million units, with output tokens almost always costing more per token tha…
The fourth installment of the "AI Foundations" series explains that a model's context window functions as short-term memory rebuilt on each request, holding the system prompt, user prompt, additional context, and convers…
Tool calling lets a language model request actions it cannot perform itself, with the application executing the tool and returning results to the model's context, according to the AI Foundations 5 explainer. Each tool re…
A guide in the AI Foundations series, "AI Foundations 6 - Prompting and Evaluation," argues that prompt structure, few-shot examples, and automated evaluation loops determine model output quality as much as the model its…
A 2026 guide compares ROCm and Vulkan as backends for hosting local LLMs on AMD GPUs, concluding that the choice depends on the inference engine, GPU generation, and workload rather than being interchangeable. It recomme…
Promptfoo 0.122.2 can be wired into GitHub Actions to run an eval suite against a golden dataset on every pull request that touches prompts, failing the CI check when outputs regress and posting a before/after comparison…
Artificial Analysis benchmarked Cognition's Devin Fusion, which pairs a frontier model with the cheaper SWE-2 model as a sidekick, finding the GPT-6 Astra pairing costs 43 percent less and runs 31 percent faster than a C…
Sakana AI released two orchestration models, Fugu Max and Fugu Ultra v2, on September 10–11, with Fugu Max priced at $2 per million input tokens and $6 per million output tokens — 40 to 60 percent below Claude Sonnet 5 a…
OpenAI's GPT-6 Astra model no longer needs the heavy scaffolding and hand-holding that earlier coding agents like Codex required, according to new guidance on rethinking skills and prompts. The guidance says skill descri…
A Thai AI practitioner's comparative analysis finds that locally trained Thai language models outperform global frontier models on Thai-specific tasks such as document and handwriting recognition, legal citation, Isan di…
Google DeepMind is reportedly near achieving Recursive Self-Improvement (RSI), according to an unverified report, with Google co-founder Sergey Brin pushing resource allocation toward RSI in Google's AI efforts around Au…
A survey of Thai-language AI resources on Hugging Face highlights underused datasets and models, including the OpenThaiGPT evaluation dataset and the WangchanThaiInstruct multi-turn conversation set, alongside SCB 10X's …