Full-text search across 68209 articles. Combine with topic and date filters; results sorted by relevance.
A measured Claude Opus 5.5 agent session of 35 calls cost $2.39 at Anthropic list prices because 97.5% of prompt tokens were cache hits, versus $22.08 without prompt caching — a 9.2x difference for identical tokens, acco…
A large-scale trace analysis of 3.7 months of production coding-agent traffic from FreeInference found that cached input tokens account for 97.4% of input-token volume and 72.9% of modeled API cost, even though the media…
A product engineer merged over 500 pull requests in about three months — roughly seven a day, with a median of 20 minutes from first commit to merge — while running several parallel Claude Code sessions in separate workt…
DeepSeek open-sourced dsh-superagent on October 31 under an MIT license, a research prototype coding agent that watches a GitHub repository, picks up authorized bug reports, writes a fix in an isolated Worker session, an…
Svelte core team member Stanislav Khromov released gpuix-svelte, an experimental project that renders Svelte components as native desktop apps for macOS, Windows and Linux without any browser or embedded webview, accordi…
Picos is a new agent harness that gives large language models a custom 32-bit CPU called smol32 with only 32 instructions (0x00–0x1f), 16 registers (r0–r15), a single flat RAM array, raw disk storage, and a 256 × 192 RGB…
Kaizen Consulting reported that its email triage system sorts about 90% of mail using roughly 750 written rules rather than model decisions, with the model consulted only when no rule matches, according to the firm's pub…
DeepSeek published dsh-superagent, an MIT-licensed research prototype coding agent for a single GitHub repository that accepts requests from GitHub and optionally Feishu and replies with pull requests. DeepSeek says the …
A course on building with Mistral models, drawing on Microsoft Foundry Models' free access to Mistral Large, Mistral Small and Mistral NeMo, advises developers to pick by constraint rather than size: Mistral Large 2 offe…
Worktrunk, an open-source Rust CLI, layers agent-oriented workflow management onto Git worktrees so developers running parallel Claude Code agents can see every worktree's status on one screen via `wt list`, with a robot…
A developer built Repo Séance, a local-first AI tool that analyzes abandoned or unfamiliar codebases by reading source trees, Git history, and Python import graphs, then answering developer questions through an entity ca…
The DevUtils MCP Server project released v1.1.0, a local Model Context Protocol server exposing 36 developer utilities — hashing, encoding, ID generation, JWT inspection, JSON formatting, network calculations and text to…
Google Cloud AI Research and Northwestern University released Harness², a framework that improves a frozen AI agent by optimizing only its harness—prompts, tools, skills, and scripts—rather than the underlying model. On …
A 3D-to-AI-video pipeline that generates a 3D model of a character with tools such as Trippo 3D, Hunyuan 3D and Meshi V7 before feeding it to AI video generators is emerging as a fix for camera-angle drift, in which AI-g…
Epoch AI reported on October 29, 2026 that Z.ai's open-weight GLM 5.3 model autonomously built a working Firefox exploit that reads a local file and sends it to an attacker, succeeding against all six releases tested (15…
A team migrated production MLflow models from Databricks Runtime 14.3 (Python 3.10) to 17.3 (Python 3.12) without retraining, using a "Load and Repack" strategy to recover cloudpickle-serialized artifacts that failed wit…
A developer built NecroShell, an open-source MIT-licensed terminal companion that hooks into Bash, PowerShell, or CMD sessions and uses a local open-weight Google Gemma model (gemma2:2b) to generate eerie commentary on c…
Cole Medin, a developer who runs an AI "second brain" built on Claude Code with Obsidian as a Markdown memory vault, reports the lean setup saves him 25–30 hours weekly, according to a first-person account of the archite…
A developer fine-tuned Qwen2.5 1.5B Instruct on self-generated agent trajectories using free Kaggle T4 GPUs, with roughly 60% of the pipeline written by DeepSeek v4.1 Flash and the rest by GLM 5.3 Flash and Mimo v2.6 Fla…
Rick Manelius launched an open-source project called Called It, built with Anthropic's Claude Fable 5.1, that grades annual predictions against three pre-written bars — strict, standard and lenient — and reports the spre…