Showing Posts From

Sr-si

Moe Hachem
November 17, 2025

The LLM Structural Crisis: Solving Context Decay with the AI Memory Prosthesis

When building complex systems with Large Language Models, I realized that the real crisis was not th...

Moe Hachem
December 10, 2025

The 200-Prompt Wall

I've spent the better part of this year building prototypes with AI assistance. Three production pro...

Moe Hachem
November 10, 2025
- AI architecture,
- Professional Practice

AI-assisted development breaks around prompt 200, but it doesn't have to

AI-assisted development breaks around prompt 200 - But it doesn't have to! Of course, your mileage...

The end of documentation debt: how an AI “memory prosthesis” automates the PRD

Product documentation has a dirty secret: it decays on contact with reality. The moment a PRD is "f...

What a brick teaches us about AI-driven product development

What does a brick have in common with AI-driven product development? Louis Kahn, one of the greates...

I shipped a beta, almost production-grade PWA fast, less than a day

I shipped a beta, almost production-grade PWA fast, less than a day. Here's the framework that made ...

Moe Hachem
February 16, 2026
- AI architecture,
- Professional practice

Spec-first is anxiety management

You're in unknown territory. Your instinct says: "We need a detailed spec because this is unfamiliar...

AI can get you 90% of the way to a spec in seconds

AI can get you 90% of the way to a spec in seconds. That might be the problem. There's something ...

Keeping UI in sync with code: Olives

What's the worst part about UI design other than doing it? Keeping it in sync with the code. Well, ...

Figma fancy screenshots vs Olives structural extraction

Figma just announced they can turn your code into fancy screenshots. Meanwhile, I've been working on...

SR-SI context savings scale progressively

AI context savings that work like taxes should: progressive. Small repos see 10-20% efficiency gains...

What if AI could learn to remember?

What if AI could learn to remember? It already does. Just not the way you think. We've been solvi...

Figma fancy screenshots vs Olives structural extraction

Figma just announced they can turn your code into fancy screenshots. Meanwhile, I've been working on...

Swap the model. The project stays coherent

The project hasn't noticed this yet but... I've been switching AI models mid-project for months. ...

The way Einstein's brain worked is how AI should retrieve information

The way Einstein's brain worked is exactly how AI should retrieve information. It doesn't. Yet. Ein...

AI governance needs identity, not just rules

Most AI governance debates are asking the wrong question. We keep asking "how do we make AI behave ...

The biggest lie in AI-assisted development

The biggest lie in AI-assisted development: "Just generate better specs and your problems will go aw...

The moat hiding in plain sight is memory

Everyone's debating whether to build up or down from general AI. Nobody's talking about the moat h...

I cut the AI's memory and it got smarter

I gave an AI a 15,800 token memory. Then I cut it to 3,300 (update: make it 1.6k). It got smarter,...

SR-SI: The methodology that gives AI persistent memory across any long-running project

106x performance improvement. A self-improving loop. And a section nobody expected to write. V2 is ...

I accidentally fixed AI forgetting without embeddings, fine-tuning, or a database

I accidentally fixed AI "forgetting" without a database, embeddings, or fine-tuning through SR-SI. ...

RAG gives AI a library. SR-SI gives it something closer to a memory

RAG gives AI a library. SR-SI gives AI something closer to a memory. The difference is smaller than...

The AI isn't an oracle, it's a collaborator

The AI isn't an oracle, it's a collaborator. There's a fundamental difference between consulting an...

Moe Hachem
February 24, 2026

The problem with AI isn't intelligence. It's orientation.

Every team I talk to has the same complaint: The outputs are generic. The AI sounds confident but ...

Moe Hachem
February 25, 2026

How I maintain coherence across 66,000 lines of code without losing the thread

Most AI-augmented development workflows break somewhere between promt 50 and 200, or as I've come to...

SR-SI accidentally solved the documentation problem too

I arrived at SR-SI to solve context decay. I accidentally solved the documentation problem too. Whe...

Moe Hachem
February 26, 2026

Why AI outputs feel 60% relevant

There's a particular type of frustrating AI output that most teams recognize right away. It’s well-s...

Computation is killing collaboration

Computation is killing collaboration. The biggest hurdle in AI-assisted development isn't a lack of...

Moe Hachem
February 27, 2026

Documentation tells you what happened. An index tells AI where to look.

Documentation tells you what happened. An index tells AI where to look. These are not the same artif...

Moe Hachem
February 28, 2026

Building four products in parallel while holding a full-time job — the system, not the hustle

This is not a productivity post. I'm not going to tell you to wake up at 5am - Because that usually ...

Moe Hachem
February 27, 2026

I don't run AI training courses. I run AI orientation engagements.

I don't run AI training courses. I run AI orientation engagements. The distinction matters more tha...

Moe Hachem
February 28, 2026

Why I went the opposite direction from spec-first

Everyone's rushing to AI tools that promise "comprehensive specs in one prompt." I went the opposit...

Moe Hachem
March 1, 2026

Why I named the studio Ocarina Systems

An ocarina is a wind instrument you can hold in one hand. No amplifier, no cables, no power source...

I've gotten lazy in the best way

People ask me: What's it like to use AI-driven development with your SR-SI and persistent memory met...

You don't fix a memory problem by adding more memory

You don't fix a memory problem by adding more memory. That's what so many AI companies are getting w...

You don't fix a memory problem by adding more memory

You don't fix a memory problem by adding more memory. That's what so many AI companies are getting w...

Moe Hachem
March 8, 2026
- AI architecture,
- Professional practice

Don't brief AI with history. Orient it with a map

You don't hand a new hire six months of meeting notes. You give them a map and help them find their ...

Moe Hachem
March 9, 2026
- AI architecture,
- Professional practice

Encyclopedia vs index: why longer prompts don't fix AI

Think about the difference between giving someone an encyclopedia and giving them an index. It's th...

What SR-SI actually is - and why I stopped explaining it as an AI trick

For the past year, whenever I described SR-SI to someone, I'd watch their eyes settle on the wrong t...

Moe Hachem
March 27, 2026

Gestalt is live. What building it revealed about the problem it solves

I built Gestalt because I kept running into the same wall. I was building multiple products simulta...

Moe Hachem
March 30, 2026

Why AI context degrades — and the architectural fix that actually works

Every team that works with AI long enough hits the same wall. The sessions start sharp. The model kn...

Moe Hachem
March 31, 2026

Protocol: building a fitness OS for coaches who've outgrown their tools

Protocol launched March 1, 2026. In its current form, it's free to use for personal training — any a...

From design intent to working components

# From Design Intent to Working Components **A codebase-first pipeline for turning visual decisions ...

Moe Hachem
March 31, 2026

From Design Intent to Working Components

Most teams solve visual consistency the same way: they rely on developers remembering the design dec...