BREAKING
06:22 UTC
Anthropic’s Claude Code source leaked via npm sourcemap; roadmap-level agent features exposed
release problems outages controversies
high
One sourcemap shipped an entire agent playbook—secure your publish pipeline and get ready to govern always-on coding agents.
anthropic
06:23 UTC
Claude Code v2.1.90: faster streaming, sturdier long sessions, onboarding “/powerup,” and tighter Windows tool permissions
new feature deep dive
medium
Upgrade to Claude Code v2.1.90 for faster streams, stabler long sessions, and safer Windows tooling, then validate resume and CI guardrails in your workflows.
cursor-ide
06:24 UTC
Cursor IDE users report severe slowdowns and regressions tied to recent builds and usage caps
release problems outages controversies
medium
Pin Cursor to a stable build and design quota-aware fallbacks; recent releases show slowdowns and breakages near plan limits.
github
06:26 UTC
Copilot goes agent-first: CLI gets CI-friendly MCP auth, Studio ships multi‑agent GA
trend pattern
medium
Agentic Copilot is getting enterprise-ready: CI-safe auth, persistent configs, and multi-agent orchestration are here—start wiring agents to your data and delivery pipelines.
chrome-devtools-mcp
06:27 UTC
Chrome DevTools MCP 0.21 ships memory-leak detection and steadier agent automation
new feature deep dive
medium
Use DevTools MCP 0.21 to put real memory and perf checks into your agents, and watch for small config and SDK quirks.
openai
06:28 UTC
Codex adds Hooks docs, community sees better limits after April 1 reset, and GPT-5.4 stop behavior raises questions
trend pattern
medium
Codex is maturing with Hooks and better limits, but double-check GPT-5.4 stop behavior before you ship.
openai
06:29 UTC
Stop Runaway LLM Agent Spend: Instrument Cost as a First-Class Metric
workflow use case
high
Treat LLM cost like latency: measure it per call, alert fast, and kill the circuit before your budget melts.
anthropic
06:31 UTC
Ship safer AI faster: put governance in CI/CD and run a model-upgrade audit
trend pattern
high
Ship faster and safer by turning AI ethics into CI tests and auditing your stack before any big model upgrade.
gemini-api
06:33 UTC
Gemini API adds OpenAI-compatible endpoint: swap three lines to try Gemini with your existing SDKs
new feature deep dive
high
You can now try Gemini with your existing OpenAI SDK by changing three lines, making multi-provider LLM setups much easier.
llamacpp
06:34 UTC
Local and edge AI cross the chasm: llama.cpp, Ollama-in-VS Code, and Akamai’s edge pitch
trend pattern
medium
Local and edge inference are ready for real workloads; start testing placement and standardizing artifacts now.
openclaw
06:35 UTC
OpenClaw buzz: China adoption claims and a push for 'free forever' local LLM setups
trend pattern
medium
Treat the OpenClaw buzz as a prompt to trial local LLM control loops—measure them on your stack before committing.
hugging-face
06:36 UTC
Code agents grow up: CI-scale benchmarking, structured patch checks, and cheaper eval runs
trend pattern
medium
Shift agent evals from single shots to CI-era maintainability, add structured patch checks, and cut costs with smaller, smarter test suites.
anthropic
06:37 UTC
Agent security is now an execution-boundary problem, not a model problem
trend pattern
high
Secure the agent’s execution boundary—tools, memory, packaging, and egress—not just the model.
massgen
06:39 UTC
From vibe coding to orchestrated agents: trace-aware memory and workflows go practical
trend pattern
medium
Agentic engineering works when you add orchestration, trace memory, and evaluators—not when you just prompt harder.
github-copilot
06:40 UTC
AI coding is an amplifier, not a shortcut—treat it as an engineering system
trend pattern
medium
AI amplifies what your process measures—so measure quality and wire those checks into the path to prod.
vector-search
06:41 UTC
Manticore Search 19.0.1 adds KNN prefiltering to HNSW for accurate, filtered vector results
new feature deep dive
medium
Apply filters inside the vector index, not after—Manticore now does this by default and your results get better fast.
anthropic
06:43 UTC
Agent Skills: A simple Markdown+YAML spec for packaging reliable agent behaviors
new feature deep dive
medium
Turn your workflows into skills: versioned, testable, and portable instructions that agents can pick and run reliably.
datasette
06:44 UTC
Datasette enrichment plugin 0.2a1 passes actor to llm.mode for better auditability
new feature deep dive
low
Actor now flows into LLM enrichments in Datasette, making audit trails and per-user logic straightforward.
datasette
06:45 UTC
Datasette-llm 0.1a6: simpler model config, better Python API docs
new feature deep dive
low
A small release that removes config duplication and clarifies the Python API, making Datasette LLM integrations a bit cleaner.