. – at Nairobi, Kenya
Replying to @_realDennis_
My workflow:
Detailed requirements doc(prd)
Derive a design system from UI inspiration
Break prd into tickets/issues
Let an agent implement them sequentially/in parallel.
A few hours can get you surprisingly far.
Then review everything...the agent will still miss things.
133
Is AI going to replace you?
Today, I had the privilege of debunking common myths, separating fact from fiction, and discussing how AI agents are transforming business processes when implemented correctly.
#DataScienceEastAfrica #DSESummit #LuxDevHQ #TechAfrica
🔒
Men at Work
At @LuxDevHQ, you get 6 to 8 months of instructor-led training and lifetime access to peers and all resources, including paperback books — all for as low as KES 30,000 (great value for your money).
We built a Hacker Hub where people push real projects into production. For those who asked about the use case of the Mini Mac, I hope you can spot it.
A practical usecase AI Agents.
Last month, @Chiiraqi and I worked on a project that showed us something simple:
Most businesses are not losing leads.
They are losing conversations that should become deals.
Here is what we found... 🧵
Are you writing code, or engineering solutions?
AI agents are great at generating code.
But they don’t fully understand the system, tradeoffs, users, or why the code should exist.
That’s where engineering matters.
Learn to think in systems.
1
67
Codex, hands down.
There are best practices you learn as you start using these tools, and even if cc is better in some ways, you won’t get to learn much if you’re hitting limits after just 3 prompts.
Also, apart from frontend work, there’s not much Opus can do that Codex can’t.
Replying to @kaka_ruto
if you have tried cc and codex, which one do you think is worth the $20?😂
I am very undecided, I have been using cursor and I would like to give a short at these two tools.
Also thanks for the insight above
174
Been stuck on a project bottleneck for 2 months . Opus 4.6 couldn’t crack it.
Let’s see if 4.7 lives up to the hype...
2
78
If you're building openclaw agents, remember to periodically run
```openclaw security audit --deep```
below is likely a case of prompt injection from a social media management skill i got from clawhub
Key takeaway from yesterday's AI agents workshop:
-> Agents work best if you treat them less like a one-off assistant and more like a teammate you configure and improve over time.
Hosting a hands on workshop today at @LuxDevHQ on building AI agents with OpenClaw
No theory here: we’re building live. You’ll leave with a working agent at the end of the workshop.
Big thanks to @HarunMbaabu, @daisyfaithauma for supporting the growing tech ecosystem



















