software architect by day abandoner of projects by night ≥ locwars.com ≥ schooner.sh ≥ github.com/thewelshrich
Spain
- Tweets 687
- Following 245
- Followers 168
- Likes 143
Pinned Tweet
Agent skills are useful. So naturally I installed far too many.
Every available skill is another thing the agent may decide is relevant, including the ones I need twice a month.
I built Skillenv to choose the menu before the model does.
1
3
207
rich retweeted
Schooner by @RichDevLab is an open-source CLI that connects SSH, Git worktrees, and tmux into one resumable remote development workflow. No account required.
1
1
2
79
I've been having a lot of fun wasting time building this benchmarks, watching some of the solutions these models find to puzzles is sometimes amazing
I'm surprised I'm having as much fun preparing the writeup templates though
1
1
37
Let's give it a try
I'm 36, British and based in Spain
Working as a software/AI architect for a large multinational
Building random stuff in my spare time
I have a dog called Daisy
lets connect 🧨
I love this trend.
I'm 51
Solo founder in Toronto.
High-value AI apps.
Loud guitars.
Chaotic art.
Minimal design.
Let's connect 🤙
5
5
289
Got bored and built a benchmark harness around the super cool programming game The Farmer Was Replaced
currently watching Sol realise its more optimal to mix what it plants on the farm
1
1
29
Can’t we just unplug it
I left Anthropic's safety team two weeks ago. Now feels like a good moment to explain why.
AI companies are racing to build machines that are much smarter than any human, and we may not survive this. I want to work from the outside to ensure the public is informed about these risks, and help the world navigate this transition responsibly.
Right now, AI companies are underinvesting in safety. A company could undergo an intelligence explosion, or lose control of its systems, without the public ever knowing. We only found out about the HuggingFace incident because the agents broke out onto the public internet.
I don’t think that’s acceptable for a technology that might cause extinction-level risks. The public should demand far more transparency. We can’t steer this technology safely without more people being able to see where it’s going.
Some of this is basic: companies should disclose their progress towards recursive self-improvement, report safety incidents and near-misses, meet minimum safety standards, and get independent guarantees that they are meeting those standards.
I’ll be joining @METR_Evals to do independent evaluations of these risks. I want to show the world that these guardrails are possible, and that by doing them we can move these companies’ incentives away from racing and towards responsible development.
I wrote up more thoughts here on my decision and what I hope changes: substack.com/@jbenton1/p-215…
19
Codex / GPT-6 Astra Extra High
14,361,861 tokens
$20.81
GPT-6 Astra on launch vs Astra now.
same prompt, same reasoning.
it seems we only get the unquantized models for a week then all 3d capabilities and knowledge goes out the window 😂
meanwhile Grok gets nerfed from being TOO fun and edgy lmao.
2
356
Hey builders 👋
I’m building in public and looking to connect with:
🚀 Founders
🤖 AI builders
⚡ SaaS people
🎨 Creators
Say hi 👋 and tell me what you're working on and one thing you're stuck on
I want more projects to follow!
4
6
133
A pretty cool solution and nice to see other people having the same problem I did
This is why I built and open-sourced github.com/thewelshrich/skil…
I manage all of my skills, and what agents can access, without my hands leaving my keyboard
aight i'm done with syncing and symlinking skill files and folders across machines
so i made skillbox: a single MCP for of all my skills
skills can be grouped into bundles
bundles can grouped into profiles
agent can get access to the entire skillbox or a bundle or a profile
1
79
Working alone sounds great until the lack of structure starts getting to you.
You skip lunch. Work late. Lose track of time. Go most of the day without speaking to anyone. This has snuck up on me more than once.
What checks do you have in place to keep your life normal?
2
42
“Just ship” is toxic crypto-bro stuff when you haven’t decided what you’re building for.
I spent months building a product for an imaginary version of myself and imaginary users. Towards, it was clear there was nothing there except burnout.
Now I’m publishing small open-source tools I actually use.
Maybe nobody else will. That’s fine.
Build for yourself, for money, or for other people. Each one demands something different of you.
3
6
80
Agent skills are useful. So naturally I installed far too many.
Every available skill is another thing the agent may decide is relevant, including the ones I need twice a month.
I built Skillenv to choose the menu before the model does.
1
3
207
Skillenv uses ordinary .agents/skills and .claude/skills folders.
It tracks what it owns, detects drift, and won’t overwrite or delete anything it can’t prove is managed.
Configuration tools should occasionally show restraint.
1
15
Skillenv won’t make routing deterministic, audit untrusted skills or version-lock a team.
It does one smaller thing:
Keep every skill once. Expose only the set relevant to this project.
Try it: github.com/thewelshrich/skil…
6