Takayama Geronimo retweeted
Your AI agent works perfectly today. Great! But what happens when it starts behaving differently tomorrow?
Dami Oshun, a Software Engineer at SeamlessHR, will be tackling this at #DevFestLagos 2026.
His session, “When AI Agents Drift: Evaluation And Monitoring In Practice With Google ADK,” will explore how to keep an eye on AI agents after deployment, evaluate their performance, and spot changes in their behaviour before they become bigger problems.
Get your ticket now! 🎟️ devfestlagos.com | 📅 Nov 13th & 14th at Lamdmark event Centre
Takayama Geronimo retweeted
PHP Talks Lagos is holding a meetup on Saturday, 31 October.
Our guest speaker, @_semeton. creator of Agent Contract, will be speaking on:
Engineering Memory for AI Coding Agents in PHP Applications
Coding agents can repeat mistakes when they lack project context. 🧵
Takayama Geronimo retweeted
Unemployment is whooping me badly! 😭
If you need a badass Product Manager or Project Manager in digital health/healthtech, I’m your girl!
Here’s my LinkedIn: ng.linkedin.com/in/aminatabo…
If you’re hiring, or know someone who is, please send a DM.
Let’s get me employed! #hireme
Repository guidance for AI Coding Agents:
open.substack.com/pub/semeto…
Since AI agents are non-deterministic, I worked on a simple system to make my coding agents work within deterministic guardrails to minimize blast radius and enable reliable outputs as much as possible (and more).
Here - agent-contract.semeton.xyz
GitHub - github.com/semeton/agent-con…
1
100
Since AI agents are non-deterministic, I worked on a simple system to make my coding agents work within deterministic guardrails to minimize blast radius and enable reliable outputs as much as possible (and more).
Here - agent-contract.semeton.xyz
GitHub - github.com/semeton/agent-con…
I have been thinking about how best to express this same thing for a while and it comforting to see someone else do justice to it. Everything to the last period is the correct diagnosis of this issue.
@falowo_ebun, remember our last episode about this.
Well, this is exactly why I don’t rate many of these people who carry philanthropy on their head like a moral crown.
You have zero interest in Nigerian politics? Very convenient.
If bad politics had not destroyed our education system, how many of these children would be depending on your philanthropy in the first place? If governance was working, how many “saviours” would still have this much relevance?
You say you care about the development of the south east, but you have no interest in the same politics that controls education, roads, jobs, security, infrastructure, business growth, and public policy in the south east.
So what exactly do you want to develop? The symptoms, while avoiding the disease?
This is the same model Nigerian pastors use. Keep people suffering, then appear with charity, prayers, scholarships, food packs, and motivational speeches. The broken system gives them relevance, so they never truly fight the system that keeps producing the victims they claim to help.
There is no difference between this type of philanthropy and Oga chess lord’s style of saviour complex. Bad governance creates the suffering, then some people build their brand from “helping” the victims of that same bad governance.
You cannot claim to love society while staying neutral about the politics destroying it.
You cannot claim to care about education while pretending politics has nothing to do with why education is dead.
You cannot claim to want the south east developed while refusing to confront the political failure that has kept it underdeveloped.
Neutrality in a collapsing country is not wisdom. It is self-preservation. And sometimes, it is business strategy.
Oga, you do not want the south east to develop. You want the south east to remain broken enough for your philanthropy to keep looking important.
It’s not poor theology is there’s no consistent way to evaluate the effectiveness of interventional prayer.
You can always invent some ex post facto justification like you did here.
The controversy surrounding this video is a function of our poor theology in this part of the world.
People cannot literally understand why someone will still die after being prayed for by a pastor!
We have projected the gospel of convenience for too long; the idea of a god that does everything as we want and not according to his perfect will.
Either we die or live, we are the Lords! A man who dies in the Lord did not loose anything! Death is nothing! Cancer is nothing! Sickness is nothing! Christ is everything!
If this prayer helped him to go through his difficult season with grace and hope, that would have been sufficient!
Brethren, Christ is our everything! We are not disadvantaged! He is always faithful, always!
3
129
Just to be clear, do you really mean this:
“That nurse might just be norm not an exception.”
I have something to say but I don’t want to make assumptions.
They are comparing a nurse not knowing the sequelae of a very common condition and being overall ignorant and clueless to a doctor unable to answer questions in WR from consultants miles ahead in experience and qualifications.
That nurse might just be norm not an exception.
55
Oh yeah, I just remembered that I did write this boring article a while back — semeton.medium.com/on-the-go…
Don’t bite me.
40
Just curious yeah, how did Messi and Ronaldo become such an icon of the sport? How do you even get to the level where you dominate your field nearly 3 decades —
24
Takayama Geronimo retweeted
Friends at both big tech and startups tell me they’re spending more than $1000 per day on Claude Code or Codex tokens.
That’s $365,000/year.
We’re not far from companies spending more on LLM tokens than on human employees.
Cursor -> Antigravity -> Copilot -> Claude -> Codex and repeat. [Copilot being the worst of all]. Tokens heavily subsidized, and soon, prices might compete with actual average* monthly SWE salaries.
Meanwhile Anthropic will turn every major SAAS product to a plugin on Claude.
2
59
We used to check 2go stars, then emails, and now Claude token usage.
So much for human evolution.
23

















