
Building growth tools for businesses. Shipping AI agents. @use_oziai |https://teapaw.org/t.co/tSwG0j0ovU | https://teapaw.org/t.co/Hi3GxLHml2 | https://teapaw.org/t.co/1tAVtCQnua
127.0.0.1
Joined July 2016
- Tweets12K
- Following1.6K
- Followers558
- Likes32K
Pinned Tweet
I wrote down how I build software as Claude Code skills. 4 plugins, 21 skills, installed once on every project.
Each one writes a real file into the repo, so the agent works from documents instead of a long chat.
How it works, start to finish 🧵
1/ Product
It interviews me in 3 short rounds, suggesting an answer to each question. Then it writes SPEC.md: requirements R1, R2… each with a pass/fail check, plus what we're NOT building.
Then PLAN.md: "T7 · Cancel booking · covers: R7 · depends: T5"
2/ Architecture
It asks at most 6 questions (scale, team, uptime, stack), works through 18 design decisions and writes ARCHITECTURE.md plus a decision record for each costly choice.
Every other skill reads that file first. On a PR, it flags code that breaks it, by file:line.
3/ AI engineering
Every model call gets a card before any code: output schema, model tier, prompt version, a fallback that doesn't use AI, and 20+ test cases.
Rules: if a regex or SQL query can do it, don't use a model. Untrusted input + a write tool = human approval.
4/ It learns
Every eval run, incident or model swap becomes a LEARNINGS.md entry with an action attached.
When the same lesson shows up in 2 projects, it asks me to promote it to the house defaults, and every new project reads those first.
5/ Agent ops
Before writing code, the agent writes GOAL.md: my request word for word, plus checks.
Every tool call is logged. A hook stops it from saying "done" until each check has fresh evidence.
Each new session opens on the current task, so nothing starts from zero.
6/ Does it work?
I test every skill against the same model without it:
• Spec from a brief: 1.0 vs 0.0
• AI design from a spec: 1.0 vs 0.0
• Goal + verify before "done": 1.0 vs 0.1
Next: security, DevOps, code standards.
github.com/ebuka-odih/agent-…
A well-written agent skill changes how the agent works, not just what it knows.
10hrs in, I've turned my engineering playbook into skills:
🏗 System design · API · data · auth
📋 Spec → plan → one task at a time
✅ Can't say "done" without proof
🔒 Security + DevOps next
github.com/ebuka-odih/agent-…
1
126
Socialism, taken to its extreme, becomes economic cannibalism. A system that gradually consumes the productive capacity of the very people expected to sustain it.
The sad part about this entire exchange is that there is a large part of the American population that will listen to what she said and tell you that it isn’t retarded and she is correct. It’s even scarier that their vote counts the same as yours
1
51
Gnosis 🥋 retweeted
You become what you do. Not what you plan. Not what you admire. Not what you talk about.
The constant pressure to steer away from mediocrity is an everyday struggle.
Sometimes, the greatest obstacle isn’t failure or doing things badly. It’s settling for what seems good but is ultimately unproductive.
You have to constantly challenge yourself, choose the difficult path, embrace growth, and remain committed to excellence.
Because staying ahead of the curve requires more than doing well. It requires refusing to become too comfortable with where you are.
12
Gnosis 🥋 retweeted
Someone asked grok about trying us. That screenshot is the reply they got
You point your agent harness at brewkeg and it just works, Claude Code, Codex, anything. Pay per token in naira, credits never expire.
-> brewkeg.dev
Gnosis 🥋 retweeted
Today we announced that LangChain raised $300M at a $3B valuation.
We think agents will transform the economy, and the companies that own their intelligence will build lasting competitive advantages.
That's why we're building LangSmith, the framework-agnostic platform to run, improve, and govern agents.
Still so much left to build! More on what’s next: langchain.com/blog/series-c
Gnosis 🥋 retweeted
This trailer by my publisher @OneSignalPub tells the story of how I founded @DAIRInstitute, which turns 5 this December!
My book DEEP UNLEARNING argues that we don't have to settle for the technological path charted by gazillionaires.
✴️Preorder at simonandschuster.com/books/D…
Gnosis 🥋 retweeted
Always a joy jamming with this wonderful players. Thank you all for listening to my version of Jolly Papa! 💜🦅
The blend of the guitar and the trumpet is simply beautiful. @WoleOjo brings this record to a close with such a lovely riff.
Thank you all for listening to my version of Jolly Papa. 💜🦅
I’m sure those in Anthropic are prolly having a champagne 🍾 right now
The fight between open/closed Ai models
Jensen being in the frontline fighting for open-weight models
Those fighting for closed Ai models will be having a good laugh right now.
11










































