I'm sharing with you one of the best papers I've read on Harness Engineering
it takes apart Claude Code (Opus 5.5), Codex, Gemini CLI and more, and boils all of them down to one line:
agent = model + harness
what you take away:
> what a harness is, in plain terms
> how modern coding agents are built under the hood
> the patterns they all share
> the direction the whole field is moving
> a checklist for building your own
the model collects the applause
the harness decides whether the agent actually works
loop and graph engineering come next, and that part lives in my article below
🚨 ALIBABA HA RESUELTO UNO DE LOS MAYORES PROBLEMAS DE LOS AGENTES DE IA.
Han creado OpenSandbox.
Un entorno seguro y aislado para que tus agentes de IA ejecuten código, naveguen por internet y controlen escritorios completos sin acceder directamente a tu PC.
¿Lo mejor? 👇
→ Ejecuta código y comandos de forma aislada
→ Permite controlar navegadores y escritorios
→ Limita el acceso a internet y otros recursos
→ Protege tus claves API con Credential Vault
→ Funciona con Docker y Kubernetes
Además, según sus pruebas, puede crear entornos aislados con Firecracker en apenas 64 ms.
Es completamente gratuito, open-source y tiene licencia Apache 2.0.
Repo 👇
DelegateEval: Evaluating Platforms for Personal External Service Agents
Jichao Li, Jinrong Lu, Manjiang Yu, Zhixuan Liu
https://t.co/xVvlhlWPjD [𝚌𝚜.𝙰𝙸]
This is f*cking insane, the CEO of Y Combinator open-sourced his own second brain and most people still don’t know it exists
8 second brain repos you haven’t seen yet:
-> gbrain (by garrytan)
https://t.co/cncTyV6CRn
the second brain of the YC CEO, every fact keeps its source and every agent shares one memory
-> llm wiki (by karpathy)
https://t.co/Z0iGVi6sI9
the original idea file behind all of this
-> ai-memory (by akitaonrails)
https://t.co/5S8nkezjq6
hand the work from Claude Code to Codex without losing context, memory lives in plain markdown
-> agent second brain (by smixs)
https://t.co/eNkyXXalyk
send a voice note in Telegram and it lands as a linked note in your Obsidian vault
-> cog second brain (by huytieu)
https://t.co/UPR34jtHqL
33 skills and 10 agents running on plain markdown, no database
-> open second brain (by itechmeat)
https://t.co/ZH1TIEGFk6
a nightly pass turns the corrections you keep repeating into rules your agent follows
-> claude second brain (by jessepinkman9900)
https://t.co/AzscisI1Z1
one npx command scaffolds Obsidian, Claude Code and local search
-> baalda (by naveedharri)
https://t.co/Brgys3RBno
a team second brain where people and agents edit the same markdown files
the CEO of Y Combinator runs his agents on this
you are still digging through bookmarks
I've seen the behind the scenes of the 4DGS input from @KakuDrop This is a REALLY solid workflow and shows what's possible with a multi camera setup & 4DGS processing combined with video diffusion workflows like using MiniMax H3 @RadianceFields
Worth checking out, it's really great work!