
OpenCode Cache Warming: What v2.2.7 Changed and When to Turn It On
OpenCode v2.2.7 makes cache warming time its pings from the model cache TTL. The config, the cost of idle pings, and when warming is a bad trade.
341 articles

OpenCode v2.3.0 makes every subagent run in the background and defaults transcripts to low verbosity. What changed, what to set, and where it bites.

OpenCode v2.2.7 makes cache warming time its pings from the model cache TTL. The config, the cost of idle pings, and when warming is a bad trade.

llama-modchip claims DeepSeek-V4-Flash 284B fits in 64GB of RAM with no GPU. What the repo says, vendor-reported speeds, what Hacker News could not reproduce, and when a hosted model wins.

Conductor is joining Meta. Cloud goes read-only Nov 13 and shuts down Nov 20, and the local Mac app winds down in early 2027. The dates, what to export, and what to move to.

OpenAI built Pando, a Git-compatible backend, after Codex flooded its monorepo with pushes. What it changed, the trade-offs, and what it means for your agent fleet.

Cloudflare's GLM-5.3 Workers AI endpoint returned other customers' data in 155 long-prompt responses. What broke in SGLang, the upgrade and config checklist, and what API users should ask.

Anthropic filed its S-1 for a Nasdaq listing under ANT. The revenue and loss figures, the risk factors that touch API users, and what to change in your stack on Monday.

CleanYourTV is a prompt plus scripts that has Claude Code strip ads and telemetry from an Android TV over Wi-Fi. What it does, where it breaks, and a safe-run checklist.

GVS5H claims five Qwen3.8-27B instances match Fable 5 on LiveCodeBench Hard. What the harness does, why Hacker News doubts the benchmark, and what to copy.

Recursive's Flint is a 40B open-weight coding model with 2.5B active parameters. How to run it on Ollama and point Claude Code at it, plus the benchmark caveats.

Google Agent Compute Engine gives every agent its own microVM with no credentials inside. What Google says it does, what is still unpublished, and what a builder can copy today.

Anthropic used Claude to prove no-crash properties for 123 OSS-Fuzz libraries. What the proofs cover, where they stop, and the fuzz-target habits worth copying on Monday.

Codex has three kinds of usage reset - banked, automatic and paid - and they move your weekly refresh date in different ways. What OpenAI documents, the 80% rule, and what to do Monday.
Showing 12 of 340 articles

New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.
Explore 847 topics
Browse All Topics