---
title: "Shafkat Rahman"
url: https://shafkatrahman.com/
description: "Shafkat Rahman — AI investment and applied AI development. Technical analysis on agent coordination, scaling laws, model architectures, and infrastructure economics."
source_sha256: d23e3aa49bf83a1b099351ed63e411cca03c6dcf9dbaf4a78f639643f4304507
---

# Shafkat Rahman

AI investment research and applied AI development.

[LinkedIn](https://www.linkedin.com/in/shafkat-rahman/ "LinkedIn") [GitHub](https://github.com/Sakeeb91 "GitHub") [X](https://x.com/Sakeeb91 "X") [Threads](https://www.threads.com/@sakeeb.rahman "Threads")

I write technical analysis on agent systems, scaling behavior, model architectures, and infrastructure economics. My background spans theoretical physics, private equity research, and shipping AI products.

## Current research

[Does the assistant persona lose stability before reward hacking?](https://shafkatrahman.com/research/persona-stability.html) Nov 2026

A pre-registered AI safety project on early warning of reward hacking. In progress, no results yet.

## Latest writing

[Software Agents Can Actually Use](https://shafkatrahman.com/writings/software-agents-can-actually-use.html) Jul 18, 2026

How to build agent-native software around shared actions, precise tool contracts, and structural security.

[The Compute Curve](https://shafkatrahman.com/writings/the-compute-curve-ai-progress.html) Jul 11, 2026

What GPU derivatives could reveal about AI progress, and why the signal will remain ambiguous.

[When Intelligence Enters the Control Loop](https://shafkatrahman.com/writings/when-intelligence-enters-control-loop.html) Jul 10, 2026

Why inference speed changes what intelligent systems can control.

## Technical analysis, by email

One useful analysis each week. No feed to keep up with.