Reuben John

[Reuben John @reubenvjohn]

Inputs: https://t.co/huVLjYcfXK
Outputs: https://t.co/UqzGaIz8iq

[Reuben John @reubenvjohn]

2/2 Result: **reposix** — a FUSE filesystem + git-remote-helper for issue trackers.

Claims **92.3% token reduction vs MCP** on the same task (~12.9× more usable context).

Only inputs I gave it: https://t.co/huVLjYcfXK

Morning brief: https://t.co/UqzGaIz8iq

[Reuben John @reubenvjohn]

Just had a "whoa, coding agents are cool" moment.

Idea: expose REST APIs (GitHub Issues, Confluence, Google Keep…) as file systems. Agents are wildly more efficient on files than tool schemas.

I told Claude Code "you have until 8 AM. Go wild. Night!"

I woke up to this 👇

[Matt Shumer @mattshumer_]

https://t.co/ivXRKXJvQg

[Reuben John @reubenvjohn]

Reeve is a prototype, but I hope to contribute its core bets to the OpenClaw roadmap:

1. Isolated sessions
2. Wrapper: Orchestrate CLIs (Goose/Claude) > Custom Runtime.
3. Desk: Git-based Memory > Vector DB.
...

Full comparison: https://t.co/StOG0z55lt
🧵3/3

[Reuben John @reubenvjohn]

🦄@steipete I💖OpenClaw! Insane!🦞
I built a similar agent (Reeve) in a vacuum—unaware of OpenClaw & "ClawHavoc" security news.
https://t.co/dbQc07TrRl
I optimized for "Session Isolation" vs. Autonomy. My research suggests this architecture fixes the current Drift issues.🧵1/3

[Reuben John @reubenvjohn]

Hypothesis: Episodic > Continuous

OpenClaw = Continuous chat (⬆️cost & context rot)
Reeve = Fresh chats

Loop: Wake (Read "Transparent" Git "Desk") ➡️ Act ➡️ Die

Benefits: • Zero Context Drift (Isolated chats) • Safety: Hallucinating? Just git revert the agent's brain. 🧵2/3

[Reuben John @reubenvjohn]

I consider myself a YouTube educational video connoisseur, but this one has been the most refreshing one I've seen in a while:
https://t.co/VbzVl12MRS

It reminds me of when I watched minutephysics for the first time.

[Reuben John @reubenvjohn]

Greetings to the 45K people on the internet who have seen this video. I suppose being a nerd is also a spectrum, and my goosebumps have revealed where I stand.
https://t.co/0IwypydaTV

[Reuben John @reubenvjohn]

Peek under the hood of one of the largest languge models: https://t.co/njVJ9k5SMI

[Reuben John @reubenvjohn]

"The Onion" - Making the world a better place... one sarcastic video at a time. https://t.co/K5AJK3QogK

[Reuben John @reubenvjohn]

It's findings such as these that remind me that Artificial General Intelligence is still a while away.
So much of our intellectual abilities are bootstrapped by evolution, not learned.
https://t.co/Wmnlx7ZW0D

#AI #AGI

[Fei-Fei Li @drfeifei]

This is the Italy I know and love!

[Reuben John @reubenvjohn]

Went Water Skiing in Kuwait last week with Dad, Richie, Adnan, Josh and Neha https://t.co/4o8M4XIzrf

[Reuben John @reubenvjohn]

https://t.co/ommx2j6FfU

[Reuben John @reubenvjohn]

Why isn't everyone talking about this! https://t.co/mc9yCkCoAt

[Reuben John @reubenvjohn]

Finally! Private contribution stats on your public profile🤓

[Reuben John @reubenvjohn]

@ideasquareio hey guys, what's going on?!

[Google DeepMind @GoogleDeepMind]

Learning by Playing - Solving Sparse Reward Tasks from Scratch: https://t.co/o74adjsA9R

"To learn complex tasks from scratch, an agent has to learn to explore and master a set of basic skills first."

[Reuben John @reubenvjohn]

@_ashish_m Feel your pain! Hoping ReiLS will save the day 😉