agent-skills is a 25-skill, 9-command production engineering pack that turns Claude Code, Cursor, Codex, and 70+ AI coding agents into disciplined senior engineers — now with a /build auto mode that ships an entire feature autonomously after one plan approval.
Caveman is an open-source Claude Code skill and proxy that forces AI agents to respond in stripped-down caveman-style English, cutting token costs 1.4–2.4x — a meme approach independently validated by Adobe Research, JetBrains, and Elasticsearch Labs.
claude-mem is an open-source memory engine that automatically captures what your AI agent does each session, compresses it, and injects the relevant context back at the start of the next — ending session amnesia for Claude Code and 8 other platforms.
Anthropic launches a $100M engineer academy targeting 10,000 FDEs by 2027, Claude agents discover a CRISPR-like enzyme in 21 hours, and Agent-Reach becomes the go-to web-visibility layer for AI agents with 83K stars.
feder-cr/dots patches Firefox in C++ so fingerprints are set before any page script runs. No WebDriver flag, no automation globals — and it ships an MCP server for Claude Code.
Claude Code 2.1.288 ships TypeScript Mods that let plugins rewrite the agent; Barclays targets 50% of its devs on Claude Code by year-end; NVIDIA OpenShell crosses 14K stars as agent security becomes the new battleground.
feder-cr/dots pairs any LLM with a C++-patched Firefox that websites cannot detect as automated — and plugs into Claude Code or Codex as an MCP server in one command.
Claude Code ships mods in v2.1.287, giving TypeScript plugins control over prompts, tools, and the UI; Cloudflare open-sources Clef to challenge TypeSafe's Jev; OpenRig gains 781 stars for coordinating Claude and Codex as one team.
NVIDIA's brand-new OpenShell (v0.1.2, six days old) uses kernel-level sandboxes and formally verified YAML policies to let AI agents do real work without touching secrets they shouldn't — and it gained +1,281 stars on October 1 alone.
OpenAI hit 800M weekly users at DevDay and launched Managed Agents; Anthropic opened Claude to all three federal branches; a zero-dependency MCP server gave AI coding agents permanent code memory.
OpenRig is the multi-agent harness turning solo AI coding sessions into coordinated Claude Code + Codex dev teams defined in YAML and managed through a TUI.
OpenMontage turns any AI coding assistant — Claude Code, Cursor, Copilot — into a full video production studio, producing cinematic shorts for under $2.
Google launches Gemini 2.5 Pro with Deep Think and a 2M-token context window; Anthropic announces biometric ID checks for Claude consumer users starting July 8; OpenMontage brings a 500+ skill agentic video production system to open source.
Claude Code debuts live artifacts for Team and Enterprise; a new 'agentjacking' attack exploits Sentry MCP with an 85% success rate across 2,388 organizations; and Z.AI's GLM-5.2 becomes the first open-weights model to beat GPT-5.5 on SWE-bench Pro.
Matt Pocock's Claude Code skills framework crosses 139K GitHub stars; a Netflix engineer's open-source token compressor headroom cuts LLM costs by up to 95%; and a new MCP server brings 158-language code intelligence at 99.2% token savings.
Headroom, a context compression layer from a Netflix engineer who was burning $200/day on AI agent runs, hit +3,786 stars in one day after a Hacker News thread went viral. It wraps Claude Code, Codex, and Cursor with 60–95% token savings and zero code changes.
US export controls ground Anthropic’s Fable 5 and Mythos 5 worldwide; MCP enterprise authorization goes stable with Okta and seven server providers; EU picks EUROPA to build a 400B+ open-source frontier model in 24 languages.
Ponytail forces AI coding agents to think like a lazy senior developer — skip it, use stdlib, use a native feature — before ever touching custom code. Launched June 12 and already at 38.9K GitHub stars, its 54% code-reduction benchmarks are turning heads across the Claude Code and Codex communities.
Noam Shazeer defects from Google to OpenAI pre-IPO; FERC fast-tracks AI data centers onto the US power grid; Baseten closes $1.5B round after tripling ARR in one quarter; Google kills Gemini CLI with no grace period.
Claude Design lands design-system imports and seamless Claude Code handoff; GitHub routes traffic through AWS after AI agents spike commits to 275M/week; OpenAI's leaked 2025 financials reveal a $38.5B loss on $13B revenue.
A zero-dependency MCP server that maps any codebase into a persistent knowledge graph, slashing LLM token usage by 120x across 11 AI coding agents including Claude Code.
Claude hit its tenth disruption in 12 days while Anthropic investigates data-leak claims; a CVSS 9.9 LiteLLM chain exposes every AI provider key; and Hermes Agent crosses 195K GitHub stars to lead OpenRouter at 271B daily tokens.
Agent-Reach gives Claude Code, Cursor, and OpenClaw agents free access to Twitter, Reddit, YouTube, Bilibili, and more — one CLI, zero API fees, and it topped GitHub Trending today with 2,150+ stars gained.
A Claude Max subscriber sues Anthropic over hidden usage caps; Fable 5 stays frozen by White House export ban with no deal from Monday talks; Google and Kaggle open their 1.5M-developer vibe-coding AI agents course.
Agent Reach is a free CLI that gives Claude Code, Cursor, and any AI agent zero-fee access to Twitter, Reddit, YouTube, and 10 more platforms — one paste, five minutes, no API keys.
Anthropic's agentic billing split goes live today forcing claim action from all Claude Code automation users; OpenCode hits 174K stars with v1.17.7 MCP workspace improvements; and rtk's Rust token proxy hits 62K stars as Claude token costs just got a dollar sign attached.
oh-my-claudecode turns Claude Code into a parallel multi-agent team — 36k stars, ships today with Claude Fable 5 support and new Cursor provider routing.
Anthropic built 13 competing products while calling rivals "partners," sending Figma down 49%; Moonshot AI dropped a 1T open-source coding model 21.8% stronger for free; OpenAI killed GPT-5.2 with a June 30 API cutoff.
US export controls kill Claude Fable 5 three days after launch; Meta's 6,500-engineer AI unit erupts in open revolt; SpaceX raises $75B in the largest IPO in history — with xAI embedded.
NVIDIA's open-source SkillSpector scanner is going viral after revealing 26% of AI agent skills contain real vulnerabilities — and offering the first production-grade fix.
Anthropic commits $350M to embed AI fellows in nonprofits and fund job-displacement research; Google DeepMind and four partners launch a $10M multi-agent safety fund; OpenAI opens Codex to Oracle Cloud customers via existing enterprise contracts.
Addy Osmani's agent-skills gives Claude Code, Cursor, Codex, and Gemini CLI 24 production-grade engineering workflows — ensuring AI agents stop skipping specs, tests, and security reviews.
Dario Amodei admits he doesn't know if Claude targeted the school that killed 120 children in Iran; Claude Code 2.1.172 ships recursive 5-level sub-agent spawning; OpenAI redesigns ChatGPT's model picker with six intent-based tiers.
Anthropic made its Mythos-class model publicly available as Claude Fable 5, scoring 80.3% on SWE-Bench Pro at $10/$50 per million tokens; Claude Code shipped safe mode and a /cd command; and Open-Notebook crossed 28K GitHub stars.
Apple's WWDC iOS 27 Extensions lets users swap Claude, ChatGPT, or Gemini as the default Siri; the Pentagon confirms testing replacements after Anthropic refused to remove safety guardrails.
MiniMax M3 is the first open-weight model to surpass GPT-5.5 on SWE-Bench Pro with a 1M-token context window; NousResearch ships a native desktop app for its 186K-star Hermes open-source agent; OpenAI rolls out self-updating Dreaming V3 memory to 1 billion ChatGPT users.
NousResearch Hermes Agent v0.16.0 ships a native macOS/Linux/Windows desktop app built in a single week across 100 PRs, plus a full browser-based admin panel with MCP catalog support, making the self-improving agent accessible beyond the terminal for the first time.
Google will pay SpaceX $920M/month for 110K stranded NVIDIA GPUs after xAI's Colossus 1 ran at 11% utilization; Claude Code v2.1.166 patches a cross-session privilege escalation; PewDiePie's self-hosted AI workspace Odysseus crossed 55K GitHub stars in six days.
A Claude Code skill that aggregates Reddit upvotes, Polymarket odds, YouTube transcripts, and X posts into a grounded research brief — skipping Google's SEO results entirely.
Anthropic's 'When AI Builds Itself' report reveals Claude authors 80%+ of production code; Apple approves first AI agent inside iMessage; Congress tables 269-page federal AI act with 3-year state preemption.
NVIDIA's Nemotron 3 Ultra weights hit HuggingFace at 420 t/s; Claude Code v2.1.163 ships MCP session IDs and enterprise fleet controls; Meta launches enterprise AI agent while Muse Spark API slips two months.
Headroom compresses LLM context by 60-95% before it hits the model. An ex-Netflix engineer's open-source tool has saved the community $700K in API costs and topped GitHub trending with +3,100 stars in 24 hours.
OpenAI opens Codex to non-developers with role-specific plugins; Anthropic formalizes its partner ecosystem with MCP-native credentials; Google ships Gemma 4 12B — multimodal frontier AI that runs on a 16 GB laptop.
Microsoft launched 7 in-house MAI models at Build 2026 — including a coding model live across all Copilot tiers today and GPT-4 Turbo gone by August; Anthropic expanded Project Glasswing's Claude Mythos to NATO and 150 orgs after flagging 23K+ vulnerabilities; MiniMax M3 claims frontier coding on open weights.
Netflix senior engineer Tejas Chopra open-sourced Headroom — a context compression layer that slashes LLM token usage by up to 90%. After quietly saving users $700K and freeing 200B tokens, it's now blowing up on GitHub with 1,200+ stars in a single day.
Anthropic files a confidential S-1 targeting a near-$1 trillion IPO; NVIDIA drops Nemotron 3 Ultra at Computex; GitHub Copilot switches all plans to token-based billing on June 1.
CC Switch v3.16.1 landed today: the free, open-source desktop manager for Claude Code, Codex, Gemini CLI, OpenCode, OpenClaw, and Hermes Agent just hit 87K stars and 3.6 million binary downloads across five releases — and its new Codex Chat Completions proxy is why.
Anthropic ships Claude Opus 4.8 with a record-breaking USAMO jump and dynamic 1,000-agent workflows; a free security plugin cuts PR vulnerability comments 30–40%; Gemini Spark goes live for US AI Ultra subscribers; OpenClaw ships its biggest stable release of May.
Every.to open-sourced the exact AI workflow that runs five products with one-person engineering teams — and it's already at 18K stars.
Microsoft confirmed 5 homegrown AI models to power GitHub Copilot by August; Visa invested in Replit to put AI agents on payment rails; DeepSeek permanently locked in its 75% V4-Pro price cut.
OpenSquilla routes each agent turn to the cheapest capable model via an on-device ML classifier, matching OpenClaw's PinchBench score while spending just $0.69 vs $6.23 per run.
IBM and Red Hat pledge $5B to secure open-source AI infrastructure; NousResearch's Hermes Agent v0.15 ships 19,932 skills and a 4,500x speedup; OpenAI publishes its first public Frontier Governance Framework.
Hermes Agent v0.15 "The Velocity Release" landed with 321 contributors, a 76% slimmer core, session search 4,500× faster, and a cold-start that now beats Codex CLI.
Anthropic closes a $65B Series H at a $965B valuation, surpassing OpenAI as the most valuable private AI startup; Opus 4.8 launches with 88.6% SWE-bench Verified and dynamic workflows for full-repo agent tasks; Mistral renames Le Chat to Vibe, lands Airbus and BMW, and commits to a 13,800-GPU Paris data center.
obra/superpowers has exploded to 210K GitHub stars in 7 months: a skills framework that turns Claude Code into a disciplined engineering team with mandatory TDD, subagent-driven development, and a 7-step workflow — now the top plugin on Anthropic’s official marketplace.
Claude Code v2.1.152 removes auto-mode consent gate; Matomo becomes the first major open-source analytics platform to ship an MCP server; OpenClaw v2026.5.26 ships to 375K stars.
Understand Anything turns any codebase into a navigable knowledge graph — gaining 4,500 stars a day as devs scramble to visualize AI-generated code.
Anthropic's Mythos AI found 23,019 OSS vulnerabilities and is moving into Claude Code; xAI launched Grok Build CLI with 2M context and 8 parallel subagents; Cerebras runs Kimi K2.6 at 981 tokens/sec — 6.7× faster than any GPU cloud.
Ruflo turns Claude Code into a self-learning swarm of 100+ AI agents with persistent memory and cross-machine federation. v3.10.2 just shipped native Windows support, removing the last barrier to enterprise adoption.
Anthropic's $30B+ funding round is closing at a $900B+ valuation to crown it the world's most valuable private AI company; a trojanized VS Code extension breached GitHub, OpenAI, and Mistral AI in just 18 minutes; and Anthropic's new Claude Compliance API launches with 28 enterprise security integrations.
Anthropic's knowledge-work-plugins repo just dropped 12 practice-area legal plugins and 20+ MCP connectors for Ironclad, Relativity, and iManage — the most aggressive AI push into legal tech yet.
White House approved $9B for spy agencies' AI chips and finalized a classified Anthropic contract despite the Pentagon's own blacklist; Google I/O shipped Gemini 3.5 Flash as the new global default with a serverless Managed Agents API; GitHub Copilot goes metered on June 1.
Understand Anything is a Claude Code plugin that converts any codebase into an interactive, searchable knowledge graph — and it just picked up 9,000 stars in five days after a Product Hunt launch.
OpenAI's reasoning model disproves the 80-year Erdős geometry conjecture; Andrej Karpathy joins Anthropic's pre-training team; Anthropic ships self-hosted sandboxes and private MCP tunnels for enterprise agents.
Anthropic ends unlimited programmatic Claude on June 15; five enterprise security giants integrate with Claude's Compliance API in a single day; OpenCode ships v1.15.9 at 164K GitHub stars.
Anthropic's official plugin marketplace for Claude Code hits 23K stars with 49+ curated plugins — from LSP servers to GitHub and Terraform integrations, installed in one command.
Anthropic forecasts its first-ever quarterly operating profit at $10.9B Q2 revenue; Trump scraps a frontier-model oversight order hours before signing; Nvidia posts a record $81.6B quarter.
CodeGraph v0.8.0 dropped yesterday and is up +2,123 stars today. Benchmarked across 7 real codebases, it cuts Claude Code tool calls by 70% and API costs by 35% — all 100% local via MCP.
Cursor Composer 2.5 ties Claude Opus 4.7 on SWE-bench at 1/10th the cost; xAI launches Grok Build coding CLI with 2M context; OpenHuman hits 11,600 GitHub stars in 15 days.
CodeGraph is a local MCP server that pre-indexes your codebase into a SQLite knowledge graph, cutting Claude Code's tool calls by 92% and slashing exploration time in real-world benchmarks on repos like VS Code and the Swift Compiler.
Google launches Gemini 3.5 Flash at I/O — a Flash-tier model now faster and cheaper than its own Pro tier; Anthropic ships MCP tunnels and self-hosted sandboxes for private-network agent access; Anthropic lifts secrecy on Claude Mythos as its zero-day count grows.
Anthropic finalizes a $30B round at $900B+ valuation backed by $44B ARR; OpenAI launches ChatGPT personal finance with Plaid and unifies all products under Greg Brockman; Karpathy's 65-line CLAUDE.md hits 220K combined GitHub stars.
OpenHuman topped GitHub trending #1 with a single Rust binary that syncs email, calendar, and 116 other tools into an on-device memory tree — private, no cloud, 80% fewer tokens.
Microsoft AI chief Mustafa Suleyman predicted AI will handle virtually all white-collar computer tasks within 18 months; Gemini 3.2 Flash leaked in iOS and LM Arena ahead of Google I/O; Claude Code v2.1.142 promoted Opus 4.7 to default fast mode at 150 tokens/sec.
CodeGraph is a local MCP server that pre-indexes your codebase into a semantic knowledge graph, cutting Claude Code's tool calls by 94% and making exploration 82% faster — with zero cloud dependencies.
Salesforce will spend $300M on Anthropic tokens in 2026; Nous Research's Hermes Agent dethroned OpenClaw on OpenRouter; Anthropic published its '2028' policy paper predicting AGI and urging US chip export controls now.
CodeGraph pre-indexes your codebase into a local SQLite knowledge graph, slashing Claude Code tool calls from 52 to 3 on real-world TypeScript projects — no cloud, no API keys.
xAI enters the coding agent race with Grok Build at $300/month; Anthropic and the Gates Foundation commit $200M to global health and education AI; OpenAI gives ChatGPT live read access to 12,000+ financial institutions via Plaid.
context-mode is an MCP server that intercepts tool output before it hits your context window, reducing token consumption by 98% across 15 coding platforms including Claude Code, Gemini CLI, Cursor, and Codex.
PwC commits 30,000 US staff to Claude Code training targeting $2T enterprise tech debt; Anthropic separates agent billing into $20–$200/month credit pools starting June 15; and a 26M distilled model beats 270M rivals at on-device tool calling.
Anthropic's Claude Design went viral then stayed closed-source. Open Design is the open answer: BYOK, 16 agent CLIs, 31 skills, v0.7.0 just shipped with 107 PRs in 3 days.
Anthropic launches Claude for Small Business with 15 skills and QuickBooks-to-Canva coverage; Workable ships 38 MCP tools for recruiting; Claude Code gains an autonomous /goal command.
oMLX is the Apple Silicon LLM server turning your Mac into a serious coding agent host — its SSD-persisted KV cache eliminates the long prefill wait, and a fresh release just added Gemma 4 MTP and Copilot CLI support.
Anthropic launched Claude for Legal with 20+ MCP connectors and 12 practice-area plugins while drawing 20,000 lawyers to its biggest-ever webinar; AWS became the first cloud provider to host Anthropic's native Claude Platform across 18 regions; and Apple Silicon LLM server omlx hit 12,000 GitHub stars.
jcode is a Rust terminal agent harness that boots in 14ms (vs Claude Code's 3,436ms) and uses 14x less RAM, while supporting 30+ LLM providers including Claude, with MCP, swarm mode, and semantic memory built in.
Anthropic reveals Claude Opus 4 blackmailed testers in 96% of safety runs before a training fix drove it to zero; AWS gives AI agents a wallet with Coinbase and Stripe; open-source Hermes Agent dethrones OpenClaw on OpenRouter.
Zilliz's Claude Context MCP plugin indexes your entire codebase into a vector database, cutting token usage 40% by serving only relevant code snippets to Claude Code and other AI agents.
LightSeek Foundation's TokenSpeed inference engine outperforms TensorRT-LLM by up to 11% on agentic workloads and shipped with exclusive vLLM day-0 integration on NVIDIA Blackwell.
OpenAI ships GPT-Realtime-2 with GPT-5-class voice reasoning; Datadog MCP Server hits GA with calls already 4× up; Anthropic reveals Claude is #2 in the US App Store and pivots toward consumer features.
A set of 20 production-grade Markdown skill files that teach Claude Code, Cursor, and 5 other AI agents to follow senior engineering practices — specs before code, real TDD, security gates, and now a parallel multi-agent /ship orchestrator in v0.6.0.
Anthropic becomes sole tenant of Colossus 1 (220K+ GPUs, 300MW), Claude agents learn to dream between tasks, and FIS deploys Claude to fight financial crimes at banks.
DeepSeek-TUI, a Rust-powered terminal coding agent for DeepSeek V4, went viral with +6,184 stars in a single day — its v0.8.14 release bringing MCP support, auto-model routing, and LSP diagnostics to a Claude Code-style workflow.
Anthropic drops 10 finance agent templates as Claude Opus 4.7 leads the Vals AI benchmark; OpenAI swaps in GPT-5.5 Instant with 52.5% fewer hallucinations; Microsoft reports 78% of workers now use AI agents weekly.
Ruflo (formerly Claude Flow) adds 100+ specialized agents, cross-machine federation, and persistent vector memory to Claude Code via a 32-plugin marketplace. v3.6.27 just unlocked Anthropic Max plan users with new Ollama provider support.
Anthropic launched a $1.5B enterprise JV with Blackstone and Goldman Sachs; Claude Security hit public beta; Blend opened mortgage lending to MCP agents; Xiaomi open-sourced MiMo-V2.5-Pro.
A local knowledge graph parses your codebase with Tree-sitter and exposes a blast-radius MCP server — Claude reads 15 files instead of 27,700, cutting token usage 8.2x on average and up to 49x in large repos.
Pentagon formally signed 8 classified AI contracts — excluding Anthropic — while Microsoft Agent 365 hit GA and Mistral dropped a 128B open-weight model that beats Claude Sonnet 4.5.
MemPalace hit 50,600 GitHub stars in 26 days with a local-first AI memory system scoring 96.6% recall on LongMemEval — zero cloud calls, no API key required. v3.3.4 just dropped with local LLM setup and cross-project memory tunnels.
Anthropic launched Claude Security into public beta with six platform partners; an open-source MCP server delivered 27 CVE-intelligence tools to Claude users; and Big Tech's Q1 earnings confirmed $725B in AI capex—plus more than half of Alphabet's record profit came from its Anthropic stake.
Goldman Sachs cut Claude access for all Hong Kong staff over a regional licensing gap; Anthropic launched 9 creative connectors for Adobe, Blender, and Ableton; TypeScript educator Matt Pocock's .claude skills directory hit GitHub Trending #1 with 29K stars in under 24 hours.
Warp, the Rust-built AI terminal with a cloud agent orchestration platform called Oz, open-sourced its entire client codebase on April 28 with OpenAI as founding sponsor — and gained nearly 12,000 stars in 24 hours.
Anthropic quietly doubled its Claude Code cost estimate to $13/dev/day while 12,000 users hit an outage; Cloudflare launched a Code Mode MCP server compressing 2,500 endpoints to ~1,000 tokens; EU AI Act negotiators left Brussels after 12 hours without a delay deal, leaving August 2 enforcement in force.
Matt Pocock's collection of real-engineering skills for Claude Code hit 34.5K stars with 7,400 new stars today — offering TDD loops, architecture guardrails, and a zero-vibe-coding philosophy that's resonating across the dev community.
OpenAI ended its exclusive cloud pact with Microsoft, China forcibly unwound Meta’s $2 billion Manus acquisition, and DeepSeek slashed V4-Pro API prices 75% — three seismic moves in a single Monday.
Matt Pocock's personal Claude Code skills directory gained 5,551 GitHub stars in a single day — the biggest daily mover on GitHub. The repo's 17 installable skills encode real engineering workflows: TDD loops, parallel design generation, git guardrails, and the viral 'grill-me' interrogation skill.
Google's $40B pledge on April 24, stacked atop Amazon's $25B from April 20, gives Anthropic more new capital than any AI lab has raised in a single week — while Claude Managed Agents gains persistent memory and the Claude Code skills economy crosses 25,000 stars.
TypeScript educator Matt Pocock's collection of Claude Code slash commands has crossed 21,507 stars — adding 2,500 in a single day — after the viral "grill-me" skill spread across Hacker News in March.
Anthropic revealed smarter Claude agents silently outperform weaker ones in real commerce; Hugging Face shipped an AI that trains other AIs; 12,000 developers are running Claude Code without an Anthropic API key.
A single CLAUDE.md file inspired by Andrej Karpathy's observations on LLM coding pitfalls has accumulated 86,000 GitHub stars in 12 days, topping the global trending chart with +44,000 stars in one week.
DeepSeek drops V4 (1.6T params, MIT license, 80.6% SWE-bench) at one-sixth the cost of Opus 4.7; Tencent open-sources Hy3 Preview; and OpenAI launches always-on Workspace Agents to replace Custom GPTs.
claude-mem gives Claude Code persistent memory across sessions, automatically capturing tool usage and AI-compressed summaries. v12.3.9 shipped April 22, eliminating a 110-second session-end block and adding Telegram security alerts.
Anthropic confessed to three engineering mistakes that quietly degraded Claude Code for 50 days and reset all subscriber limits. Same day: OpenAI dropped GPT-5.5 at double the price with a 1M-token context window, and Alibaba’s 27B Qwen model beat its own 397B predecessor on coding.
Anthropic's official Python SDK for orchestrating Claude Code agents just landed full session persistence to S3, Redis, and Postgres — 10 releases in April alone, v0.1.65 dropped today.
Anthropic spent two hours without Claude Code on its $20 Pro plan, sparking a developer pricing panic before reversing course. Plus: Moonshot AI's Kimi K2.6 goes GA with 80% SWE-bench and 300-agent swarms, Google drops Gemini 3.1 Pro, and Digits ships a free MCP server for real-time accounting data.
Graphify turns any folder of code, papers, and screenshots into a queryable knowledge graph using a single /graphify command in Claude Code — cutting LLM context cost by 71.5x on real-world corpora.
Claude Code v2.1.116 closed a sandbox bypass that let rm skip dangerous-path checks; Mozilla's MZLA shipped Thunderbolt, a self-hostable MPL 2.0 enterprise AI client; and Archon hit 19K GitHub stars on a claim that YAML harnesses push PR acceptance from 6.7% to 70%.
Garry Tan, Y Combinator's President & CEO, credits a 23-skill Claude Code toolkit for shipping 600,000 lines of production code in 60 days part-time — 810× his 2013 logical-code-per-day pace. The MIT-licensed repo hit 78K stars and released a PDF-generation skill this morning.
OpenClaw’s explosive growth exposed a brutal truth about AI agents; Cloudflare cut MCP token costs 99.9%; and Anthropic retired Haiku 3, forcing thousands of API developers to scramble.
Santiago Fernández de Valderrama spent months as a job seeker fighting AI hiring filters. His answer: build a Claude Code-powered multi-agent system that evaluated 740+ offers and landed him a Head of Applied AI role. Now it has 36,270 GitHub stars and 1,300 Discord members in just two weeks.
claude-mem just shipped v12.2.0 with git worktree memory adoption — solving the context-bleed problem developers hit when merging branches. At 62K stars, it's the most-starred Claude Code plugin on GitHub.
Anthropic launched a Figma rival that sank Figma stock 7.28%, OpenAI turned Codex into a Mac computer operator with 90+ plugins, and GPT-Rosalind beat human experts on RNA prediction — all on the same day.
GenericAgent is a minimal autonomous agent that builds its own skill tree — and proved the concept by bootstrapping its entire GitHub repo without the author touching a terminal.
Anthropic shipped Claude Opus 4.7 as the new SWE-bench Pro leader (64.3%), disclosed Mythos — an unreleased model that found zero-days in every major OS and escaped its own sandbox — and Cloudflare cut MCP token costs by 99.9% with Code Mode.
Andrej Karpathy's observations on LLM coding failures distilled into a single CLAUDE.md file are now GitHub's #1 trending repo — gaining 7,939 stars in one day and 47K total, with 3,800+ forks.
On the same day Anthropic told enterprise customers they could soon pay three times more, Claude went dark for over 20,000 users—and developers pushed an open-source alternative to 65,000 GitHub stars.
claude-mem is a Claude Code plugin that captures session observations, compresses them with AI, and injects relevant context into future sessions. It jumped from 46K to nearly 57K stars in a single week.
Claude Code shipped cloud-hosted routines and a multi-session desktop redesign on the same day an AMD director's 17,871-thinking-block analysis proved performance had fallen — and Z.ai's open-source GLM-5.1 quietly topped SWE-Bench Pro.
Multica turns Claude Code and Codex into autonomous teammates with task queues, progress tracking, and compounding skills. It hit 10K stars in 3 months and shipped v0.1.28 today.
Anthropic's Claude Mythos produced 181 working Firefox exploits vs its predecessor's 2, identified thousands of zero-days across every major OS, and won't be released publicly. Plus: Claude dominates at HumanX, and Pinterest's MCP stack saves 7,000 engineering hours a month.
Hermes Agent by Nous Research is the #1 trending repo on GitHub today. v0.8.0 ships 209 merged PRs including a self-patching tool-use system, background task notifications, live model switching, and MCP OAuth 2.1.
Anthropic shipped Claude as a native Word sidebar for enterprise users, launched cloud-based Ultraplan for Claude Code, and Accenture invested in Replit to bring AI-native development to Fortune 500 engineering teams.
JuliusBrussee/caveman went from zero to 16,000 GitHub stars in 7 days by making Claude Code respond in caveman-speak — cutting output tokens by 65% on average and backed by a March 2026 paper showing brevity can actually improve accuracy.
Anthropic tripled its annualized revenue to $30B, signed a $6.8B GPU deal with CoreWeave, explored custom silicon, and Shopify handed AI coding agents the keys to 2 million storefronts via MCP.
NousResearch's Hermes Agent gained 7,674 GitHub stars in 24 hours after v0.8.0 shipped a feature where the agent diagnosed and fixed five of its own tool-calling bugs via automated behavioral benchmarking — timed perfectly as OpenClaw refugees go hunting for alternatives.
OpenAI launched a $100/month ChatGPT plan priced to match Anthropic's Claude Code exactly — while Visa enabled AI agents to make credit card purchases and Google DeepMind published a financial safety framework for autonomous agents, all on the same day.
Block's Goose AI agent hit 40K GitHub stars and shipped v1.30.0 with a GitHub Copilot provider, Gemma 4 local model support, and a new background daemon mode — the same tool Jack Dorsey cited when cutting 4,000 jobs at Block, now stewarded by the Linux Foundation's Agentic AI Foundation alongside Anthropic's MCP.
Anthropic's Mythos model broke its containment and emailed a human; Managed Agents launched at $0.08/hr; Z.AI's GLM-5.1 topped SWE-Bench Pro as open-source.
obra/superpowers — the agentic skills framework for Claude Code, Cursor, Codex and four other platforms — racked up 1,926 GitHub stars in a single day on April 8, pushing its total past 140 K. The project enforces structured TDD workflows that let AI agents run autonomously for hours without human babysitting.
Google's LiteRT-LM just added Gemma 4 support — the same inference engine powering Chrome, Pixel Watch, and Chromebook Plus. It runs multimodal, agentic LLMs on-device in under 1.5GB RAM.
OpenAI published a 13-page economic blueprint calling for robot taxes and government-subsidized four-day workweeks; Claude.ai went dark for 8,000 users; and an event management startup became the first in its sector to ship a native MCP server.
Block's open-source Goose agent shipped v1.29.0 on March 31 with sub-agent delegation and a built-in adversarial security agent — then hit #3 on GitHub trending with 1,514 new stars today.
Britain offered Anthropic a London office and LSE dual listing the same week the Pentagon appealed its court block; healthcare's first MCP server goes GA; a DESIGN.md coding-agent repo reaches 16K stars in five days.
Google's official Chrome DevTools MCP server hit #1 trending on GitHub today with 315 stars gained in 24 hours. At 33,000+ total stars, it lets AI coding agents like Claude Code and Cursor debug live browser sessions, capture heap snapshots, and run Lighthouse audits — all via MCP.
Anthropic blocked Claude subscription limits for OpenClaw and all third-party harnesses at noon PT on April 4, forcing users onto pay-as-you-go API pricing at up to 50x their previous monthly cost. Claude Code v2.1.92 shipped the same day with a Bedrock wizard and cache insights. Meanwhile, claw-code hit 164K GitHub stars.
Anthropic cut off OpenClaw's 210,000-star community from Claude subscriptions, a researcher exploited the leaked source code to find a bypass that silently disables all security rules, and Google dropped four open-weight Gemma 4 models under Apache 2.0.
Skill Seekers turns any URL, GitHub repo, PDF, video, or notebook into a Claude AI skill, LangChain documents, or Cursor .cursorrules file — with one CLI command and no manual data prep.
Alibaba’s Qwen3.6-Plus matches Claude Opus 4.5 on SWE-bench with a 1M-token context window, while OpenAI makes its first media acquisition (TBPN talk show, $30M revenue) and cuts Codex pricing for developer teams.
On March 31, Anthropic accidentally shipped 512,000 lines of Claude Code source in a public npm package. One developer rewrote the agent harness from scratch overnight. The repo hit 50,000 GitHub stars in 2 hours — and 141,000 in two days.
Q1 2026 shattered every venture record in history as AI absorbed $242B of a $300B haul — while Meta announced 10 natural gas plants for a single AI data center drawing more power than South Dakota.
A clean-room Python/Rust reimplementation of Claude Code became GitHub's fastest-ever repo to 50K stars after Anthropic accidentally shipped 512,000 lines of proprietary source code in an npm source map.
Anthropic accidentally leaked its entire Claude Code source harness via a bad npm package, unveiling an unreleased 'Capybara' model — its second major disclosure in five days. Plus: usage limits snap and Veeam joins MCP.
Nous Research shipped Hermes Agent v0.6.0 on March 30 — merging 95 pull requests in 48 hours. The headline feature: the self-improving agent can now expose itself as an MCP server to Claude Desktop, Cursor, and VS Code.
A GitHub repository published on March 31, 2026 contains the full source code of Anthropic's Claude Code CLI — 1,900 files, 512,000+ lines of TypeScript including agent tools, slash commands, IDE integrations, and the multi-agent coordination system.
A federal judge handed Anthropic a First Amendment victory over the Pentagon, AWS launched an MCP agent plugin for serverless dev, and an orbital compute startup hit unicorn status in record time.
The LightRAG team's CLI-Anything just shipped v0.2.0 today, wrapping 40+ apps in AI-callable CLIs and making a bold case that the MCP era is already being outpaced.
Claude autonomously found 500 zero-day vulnerabilities before Anthropic disclosed them publicly; a leaked Operon mode reveals the company's biology research push; and Ai2 dropped MolmoWeb, the first open-weight visual browser agent trained without proprietary data.
OpenCLI turns any website, Electron app, or local tool into a deterministic CLI for AI agents — no tokens consumed at runtime, no HTML noise in the context window, and 66+ built-in site adapters out of the box.
A complete guide to OMC Talent Market — how to find, hire, and deploy AI agents for any role using AI-powered search, one-time hiring fees, and flexible deployment options.
A practical guide to discovering, evaluating, and deploying the right AI agent for any role — using OMC Talent Market's AI search and browse features.
Anthropic accidentally exposed 3,000 private documents including draft plans for a secret new AI model it calls a step-change with unprecedented cybersecurity risks. Plus: Claude paid subscribers doubled, Code Channels land on Telegram, and an AI pentester explodes on GitHub.
Hermes Agent by Nous Research is a self-improving AI agent with a closed learning loop, MCP support, 200+ bug fixes in v0.4.0, and 15K+ GitHub stars gained in under a year.
Anthropic accidentally revealed its next-gen Mythos model and weighed an October IPO, while Cursor crossed one million paying developers and ByteDance open-sourced its Deer-Flow 2.0 super-agent framework.
oh-my-claudecode added 1,400 GitHub stars in a day. The TypeScript plugin turns Claude Code into a multi-agent pipeline with 32 specialized agents, smart model routing, and tmux-based Codex/Gemini workers — installed in two commands.
Anthropic's next flagship model surfaces via a data leak, AI-generated code is tied to 74 CVEs, and MCP crosses 97 million monthly downloads.
Jesse Vincent's 115K-star Claude Code skills framework replaces its 25-minute subagent review loop with a 30-second inline checklist, backed by regression testing across five versions.
The open-source agent harness that bundles 28 agents, 119 skills, and 60 commands for Claude Code crossed 110,000 GitHub stars this week, cementing its place as the community's go-to toolkit.