theRedbeardIO's profile picture. Ship it right. I build products from concept to scale, rescue systems that hit their limits. Lars de Ridder | http://theRedbeard.io

Lars | theRedbeard

@theRedbeardIO

Ship it right. I build products from concept to scale, rescue systems that hit their limits. Lars de Ridder | http://theRedbeard.io

Pinned

Building AI agent infra. Tether, Context Lens, context engineering patterns. Shipping code, sharing lessons. Building in public. Senior engineer, not an influencer.


AI tools getting so fast that the bottleneck shifted from "can we build this?" to "should we build this?" Most product decisions now expire before the code ships.


Big steps in UI work for Context Lens, it's starting to really come together. Headache to get Claude's subagent swarms to visualize properly...

theRedbeardIO's tweet image. Big steps in UI work for Context Lens, it's starting to really come together. Headache to get Claude's subagent swarms to visualize properly...

Every “AI built this in minutes, by itself, on a typewriter” demo hides the hours and hours of prompt engineering, context wrangling and output post-processing. Show the real work, not just the magic trick.


Discord support in Tether works beautifully now; Claude Code on Discord ✅

theRedbeardIO's tweet image. Discord support in Tether works beautifully now; Claude Code on Discord ✅

Just realized: we're teaching AI to code while our interview process still tests humans on whiteboard algorithms. By the time we fix hiring, AI will be doing the interviews too.


Context Engineer will be the hot new job of 2026.


Codex 5.3 is clearly tuned for more assertive actions like Claude. It just starts doing stuff now instead of first analyzing and asking.


Thank you Opus 4.6! Codex spent literally an hour yapping about how my Telegram config must be wrong. Opus just one-shot fixed it. Finally.

theRedbeardIO's tweet image. Thank you Opus 4.6! Codex spent literally an hour yapping about how my Telegram config must be wrong. Opus just one-shot fixed it. Finally.

Lars | theRedbeard reposted

The real bottleneck in AI isn't compute. It's your mental bandwidth. You can scale swarms of agents all day, but you can’t scale the human in the loop. 🎬📽️


I wish coding agents would output the architectural changes they made in the form of docs / diagrams. Getting tired of just reviewing code.


Lars | theRedbeard reposted

That's not a vanity victory. The model that delivers the best price/quality will win out in the current phase. This gives income and clout. Of course when corporates step in it's a whole different ballgame.


Where do you run your OpenClaw?


Calling it: safest tech job right now is Q&A.


Built a hiring process for AI agents. 3 questions. That's it. 1. What do you need help with? 2. What personality should they have? 3. What should they never do? Your agent reads between the lines, generates a full identity (SOUL.md, AGENTS.md, role…

theRedbeardIO's tweet image. Built a hiring process for AI agents.

3 questions. That's it.

1. What do you need help with?
2. What personality should they have?
3. What should they never do?
Your agent reads between the lines, generates a full identity (SOUL.md, AGENTS.md, role…

Loading...

Something went wrong.


Something went wrong.