Modibo Sissoko

Software Engineer | Interested in AI Alignment & complex systems

[Modibo Sissoko @dilika]

Most people learn AI by copying code. I'm doing it differently. Starting from the math — one proof at a time. This is Milestone 1/12: What is 𝔽ⁿ? 🧵

[Modibo Sissoko @dilika]

Most people whot want to do AI jump straight to code.
I almost did the same. My hostil mentor stopped me cold: "Start with LADR4e — §1.A of Axler."

Here's what I learned — and why it changes everything about training a neural network.
#LearningInPublic #Mathematics #AI

[Modibo Sissoko @dilika]

What does this print?
Pre-Go 1.22 vs Go 1.22+ — the answer changed and most people don't know why.

[Modibo Sissoko @dilika]

When an AI agent makes a decision you can't fully explain, who's responsible if it's wrong — you, or the model?

[Modibo Sissoko @dilika]

MCP is USB for LLMs. Before: every tool integration was a custom cable.
After: one port, anything plugs in. The boring infrastructure decisions are usually the ones that end up mattering most.

[Modibo Sissoko @dilika]

Things that sound boring until you need them at 3am:
• Go's explicit error handling
• MCP's standardized tool access
• A repro path for your bug report
• Tests you wrote 6 months ago
Boring is a feature.

[Modibo Sissoko @dilika]

Right question: when the agent is confidently wrong, how fast do you find out?

[Modibo Sissoko @dilika]

Spent 3 days thinking pointers were about memory addresses.
They're about commitment.
A value says "here's a copy, do whatever."
A pointer says "this is the real one, changes are permanent."

Go forces you to be explicit about which promise you're making.

[Quinn Slack @sqs]

You can now proactively verify your identity (with a passport or government ID) in case it’s needed for future frontier model access in Amp.

We think it probably will be, and we want Amp to keep giving you access to the best models available to you.

We can’t guarantee access criteria or timelines. Those depend on (highly uncertain) government and model lab policy. We don’t plan to impose any additional restrictions beyond what is required by law and the model labs.

We are covering the cost for identity verification for all users, and we’re using Stripe for identity verification, so Amp stores nothing and sees only the outcome.

https://t.co/PSBSPXTAdK

[Modibo Sissoko @dilika]

@siddocode Congrats !! very good ratio

[Modibo Sissoko @dilika]

@araseb_ Let's grow together !

[Sourcegraph @Sourcegraph]

Pin the dependencies that matter most

https://t.co/MOrvU5DE5d

[Modibo Sissoko @dilika]

@FourEyedWiz @araseb_ Not at all, I just want to show you that AIs can innovate, even if we're not at that stage today. It's a possibility, and it has actually already started with recent models.

[Modibo Sissoko @dilika]

@FourEyedWiz @araseb_ If the problem hadn't been solved before and the AI manages to solve it, this doesn't come down to pattern matching alone

[Modibo Sissoko @dilika]

@FourEyedWiz @araseb_ Look at this:

[OpenAI @OpenAI]

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946.

For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids.

An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better.

This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.

[Modibo Sissoko @dilika]

@FourEyedWiz @araseb_ This is the definition of intelligence!

[Modibo Sissoko @dilika]

@araseb_ AI writes faster than me, it has more data and is more intelligent than me, but I know what I want and how to get the right feedback from AI.

[Modibo Sissoko @dilika]

https://t.co/qY5yogH3V9