Category: tech > ai

Claude Code's /goal command runs an autonomous loop: take an action, check the goal condition, repeat until met, replacing step-by-step prompting.

A sharp-opinion FAQ on LLM evals: skip generic metrics and tooling, do error analysis on your own traces, build a failure taxonomy, then write targeted evalu...

Reflexion lets an LLM agent learn from failures without fine-tuning, writing a verbal post-mortem and feeding it back as context, lifting GPT-4 to 91% on Hum...

A short clip claims Claude Code now auto-generates a Live Artifact summarizing code and PRs at the end of a session to share with teammates.

A short teaser for loop engineering, the idea that AI coding agents work best run in an iterative loop rather than expecting one-shot results.

Open-source (Apache-2.0) foundation speech-recognition model built for messy audio, claiming up to ~30% WER gains over SOTA where other models collapse.

Local Qwen 27B isn't near-Opus for coding, but it pays off on privacy-bound analysis where cloud models are contractually off-limits.

Open-source MIT orchestrator that drives coding agents through a structured plan, running each task in a fresh session plus multi-agent code review.

Anthropic launched Claude Fable 5, a Mythos-class model with request-screening safeguards, built for long-running autonomous work beyond coding.

Loop engineering is an agentic workflow where you set one goal and let agents self-prompt through discovery, planning, parallel execution, and verification.

Claude Code Dynamic Workflows runs hundreds of parallel sub-agents in one session, compressing multi-day coding into a day at much higher token cost.

Claude Code Workflows lets a workflow.js script invoke subagents in loops, sequences, or parallel, with structured output schemas to chain agents reliably.

Dozens of undocumented Claude Code config fields hidden in the npm source: command-rewriting hooks, persistent agent memory, and a plain-English auto-approva...

Google's Gemma 4 12B is an open Apache 2.0 model handling text, vision, and audio with no separate encoders, fitting in 16 GB while nearing 26B benchmarks.

Lathe is a Go CLI and LLM skill system that generates multi-part hands-on tutorials on demand, then serves them locally so you do the work by hand.

A single-binary Go coding agent that runs with no runtime dependencies. Provider-agnostic with intentional minimalism: four core tools and an extension model.

An MIT-licensed, terminal-first AI coding agent built only for DeepSeek, engineered around its prefix-cache to keep token costs low across long sessions.

Community directory cataloguing every connector in Anthropic's official Claude Connectors Directory, 414 verified MCP servers across 30 categories.

Google's autonomous coding agent takes a GitHub repo and a prompt, plans with Gemini 3 Pro, then produces a reviewable diff and opens a PR.

AI lowers the bar to build software, but engineers shift from writing code to managing agent teams and owning quality. Junior hiring is down nearly 20% since...