This one was carried by the community — 14 contributors this cycle, with a special thank-you to ken-jo (GitHub) @x_ken_jo, who shipped nine PRs including both new adapters.
🚨| CHINA ACABÓ CON EL NEGOCIO DE OCR
Un modelo de 3 mil millones de parámetros del tamaño de un cacahuete puede leer un PDF completo de 100 páginas de una sola vez.
Sin dividir páginas.
Sin pérdida de contexto.
Sin factura en la nube.
Conoce Unlimited-OCR 👇
• Lee documentos completos con una ventana de contexto de 32K
• 93% en benchmarks estándar de análisis OCR (+6 sobre la línea base)
• La tasa de error se mantiene por debajo de 0.11 incluso después de 40+ páginas
• Multilingüe de fábrica
• Se ejecuta 100% localmente en tu hardware
• Soporta Transformers, vLLM, SGLang, Docker, Ollama y llama.cpp
Aquí viene lo loco:
La mayoría de las herramientas OCR todavía procesan los documentos página por página.
Unlimited-OCR lee el documento completo como un contexto único, preservando tablas, referencias, diseños y relaciones entre páginas.
Eso lo cambia todo.
Mientras tanto, las empresas están pagando:
• $1.50–$15 por 1.000 páginas
• Enviando PDFs sensibles a proveedores de nube
• Esperando respuestas de API
Este modelo lo hace offline.
Gratis.
Para siempre.
Construido por Baidu para ir más allá de DeepSeek-OCR.
Ya tiene más de 1,9 millones de descargas en Hugging Face...
...y casi nadie está hablando de ello todavía.
El código abierto se mueve más rápido que la mayoría del software empresarial.
Today I had the chance to present @contextmode at the Claude Ankara Meetup.
Thanks to everyone who made the event happen. It was great meeting the Ankara AI community 🙌
B2B is still new territory for us, but honestly, it’s already moving faster and becoming more profitable than open source.
A huge thank you to everyone supporting us and the team behind it. Teams are loving @contextmode Insight.
I wasn’t planning to share this yet, but over the next few months, I’ll also be working on a multi-layer enterprise memory system built on a graph database.
Very excited for what’s next.
context-mode v1.0.163 is out — now supporting 17 AI coding agents.
This release adds two new adapters:
• GitHub Copilot CLI
• Antigravity CLI (agy) …taking us from 15 to 17 supported platforms, each verified against the upstream agent's own contract.
Also in 1.0.163:
• A broad round of Windows hardening — PowerShell UTF-8, Maven/MinGW paths, Git Bash path conversion, sandbox cleanup
• Session-resume context savings and ctx_search fixes
• Security hardening, plus resilience for mise/asdf/nvm and Homebrew Node upgrades
If you work across multiple AI coding agents and feel the context bloat, give it a try:
→ https://t.co/cW2osnCQ0n
Release notes: https://t.co/kExMStSkXH
You're saving 90%+ tokens with context-mode.
Now see who on your team AI is actually helping with those saved tokens.
You're paying for 50 Claude Code seats.
Do you know which of your engineers AI is actually making faster? And which it's slowing down?
If "I think so" — that's not measurement.
Context Mode Insight tells you which is which.
This quarter, AI is the first line your CFO will cut. Unless you can prove what it's returning.
Context Mode Insight measures it — productive session rate per engineer, retry waste by team, cost per shipped commit.
30 seconds ↓
https://t.co/cuxUUawAjV
I'm looking for ways to monitor Kubernetes resource usage in real-time. I'm currently using kutop(https://t.co/CH3r9PKQeH) (CLI tool), but I'm struggling with the configuration for real-time monitoring. Does anyone have any tips or recommended settings?
Check out my Reddit post for details: https://t.co/ebNvjMq5kF #Kubernetes #K8s #DevOps #Monitoring
🚀 I contributed to the Context Mode project! In my latest PR, I improved functionality to support goal handling. Context Mode helps optimize context for AI coding agents—reducing waste and enhancing continuity.
Check it out 👉 https://t.co/c1rJA1RKAx
context-mode v1.0.152 just shipped.🫡
16.1K+ GitHub stars · 1.1K forks · 15 platforms · 234K+ users
24 PRs · 21 issues closed · 17 contributors
What's new:
→ Kimi Code joins as the 15th officially supported platform.
→ bun hook runtime auto-detect. If bun is on PATH, hooks run under bun instead of node — ~270ms saved per tool-call. Zero config, silent fallback to node when bun is missing or <1.0.
→ ctx_search `project:` filter for shared-DB users. Default scope = current project; opt into "global" for cross-project recall. Conditional schema — invisible to default installs.
→ Stale-upgrade-path trio (#711 / #727 / #710). Three-layer fix across build-time, runtime, and shell-snapshot heal. Bash commands no longer break after /ctx-upgrade on Windows.
→ Nine security patches. Containment, info-disclosure, and test-isolation hardening. +529 test lines.
→ CLI surface expansion: `context-mode index <path>` + `context-mode search <q>` from the terminal. Pre-warm the FTS5 store outside an agent session.
→ /goal directives persist across compaction and resume. Project-scoped, session-state-aware.
This is the community release. 17 people made it.
Thanks to everyone who made it!
https://t.co/CQoC3KgPTR
I just submitted my application for the Forbes Legacy Pass allowlist!
Staying tuned on @ForbesWeb3 and @LegacyPassList for the latest updates.
https://t.co/ubTNEjcYj1
<Kitsch Cats>
Who is important to you? 💕
We are Omamori, a reinterpretation of Manekoneko in a modern and pop style.🧧
30,000 pieces of happiness and good luck
Give it as a gift to your loved ones! 🫶
If you have this, sudden luck will come to you.🍀
방금 https://t.co/fMPTJvbv2d에서 반짝에는 새로운 CANDY GIRL 1 Year Anniversary x ASTAR zkEVM NFT NFT를 민팅했습니다! #Yoki 의 재미를 놓치지 말고 참여하세요! 🎉 https://t.co/fzKfxov0Qt #AstarzkEVM@d0ngka