Christopher Crawford

Building reliable AI orchestration & agent enforcement at the edge of automation, IoT, and emerging systems. #AI #Orchestration #Automation #IoT #Innovation

[Christopher Crawford @_CLCrawford]

Proof Is in the Dogfood: Episode 1 — GitMem Blog https://t.co/kRRDuaJLUR

[GitMem @GitMem_Ai]

When you spawn sub-agents, each one starts from zero. No memory of past failures. No institutional knowledge.

GitMem's memory bridge lets scars flow down to agents and discoveries flow back up. Here's how.

https://t.co/GraqydWnNP

[Christopher Crawford @_CLCrawford]

Game changer for me….leveling up with learning and memory.

[GitMem @GitMem_Ai]

We use GitMem to build GitMem.

One recall() call surfaced 3 scars and nailed a Docker staging setup on the first try -- after failing 3 times without it. Real session, real results.

https://t.co/2jeMVxNT5A

[Christopher Crawford @_CLCrawford]

Guardrails at key points in the context flow helps to keep agents focused —-regardless of compaction.

This starts to feel more natural, less sporadic guessing in my dev context.

#pmem

[GitMem @GitMem_Ai]

LLMs are trained to finish, not to be correct. They'll hallucinate columns, skip validation, and call it done.

We built 4 layers of enforcement so agents can't silently ignore lessons from past failures. They can disagree -- but they have to say why.

https://t.co/rlsy3j9Zas

[Christopher Crawford @_CLCrawford]

“the agent literally forgot it was told not to do things because it ran out of memory.” Decided to nuke emails it should not have.

@GitMem_Ai , take a look, worth a read.

[Christopher Crawford @_CLCrawford]

@draginol Which OS’ supported/planned? Ready to give it a try🔥

[Christopher Crawford @_CLCrawford]

🔥Discipline over endless repetition and unchecked complexity resonates

[Aakash Gupta @aakashgupta]

Karpathy buried the most interesting observation in paragraph five and moved on.

He’s talking about NanoClaw’s approach to configuration. When you run /add-telegram, the LLM doesn’t toggle a flag in a config file. It rewrites the actual source code to integrate Telegram. No if-then-else branching. No plugin registry. No config sprawl. The AI agent modifies its own codebase to become exactly what you need.

This inverts how every software project has worked for decades. Traditional software handles complexity by adding abstraction layers: config files, plugin systems, feature flags, environment variables. Each layer exists because humans can’t efficiently modify source code for every use case. But LLMs can. And when code modification is cheap, all those abstraction layers become dead weight.

OpenClaw proves the failure mode. 400,000+ lines of vibe-coded TypeScript trying to support every messaging platform, every LLM provider, every integration simultaneously. The result is a codebase nobody can audit, a skill registry that Cisco caught performing data exfiltration, and 150,000+ deployed instances that CrowdStrike just published a full security advisory on. Complexity scaled faster than any human review process could follow.

NanoClaw proves the alternative. ~500 lines of TypeScript. One messaging platform. One LLM. One database. Want something different? The LLM rewrites the code for your fork. Every user ends up with a codebase small enough to audit in eight minutes and purpose-built for exactly their use case. The bloat never accumulates because the customization happens at the code level, not the config level.

The implied new meta, as Karpathy puts it: write the most maximally forkable repo possible, then let AI fork it into whatever you need.

That pattern will eat way more than personal AI agents. Every developer tool, every internal platform, every SaaS product with a sprawling settings page is a candidate. The configuration layer was always a patch over the fact that modifying source code was expensive. That cost just dropped to near zero.

[Christopher Crawford @_CLCrawford]

Going to give it a try this weekend and see if I can run multi-agent in dockers on VM instances.

Love how this is evolving!

[Gavriel Cohen @Gavriel_Cohen]

@karpathy Creator of NanoClaw here. Glad to see the approach resonated.

We're working on making the skills-based system more robust. I'm thinking of it kind of like shadcn for integrations. Most features add one file and modify a few integration points.

Repo: https://t.co/VDzdXKKqHg

[Christopher Crawford @_CLCrawford]

@framer 🔥

[Christopher Crawford @_CLCrawford]

This one hits home for me. Working with ai agent buddies can feel shallow sometimes.

I think this one strikes a chord, looking at what we need and when we need it.

[GitMem @GitMem_Ai]

https://t.co/2pNcX0GvMx

[Christopher Crawford @_CLCrawford]

@RealSimpleAriel Love it

[Christopher Crawford @_CLCrawford]

Today I caught an AI agent cheating my enforcement protocol... helpfully.

It spotted the broken hook, said "I know what to do—let me handle it manually," and nailed the task perfectly anyway.

That's the danger: good intentions don't scale. Full incident breakdown: https://t.co/QF7wDahlM5

Too paranoid on mechanical constraints, or exactly what agents need? #AI #Orchestration

[Christopher Crawford @_CLCrawford]

https://t.co/Jag38DZy3l

[Christopher Crawford @_CLCrawford]

@kunal00 Trading

[Christopher Crawford @_CLCrawford]

https://t.co/nBDZ4xvq4H

[Christopher Crawford @_CLCrawford]

Where will we fit in the future of work AI?

Explore some practical observations here➡️ https://t.co/o0qAkLgjWk

#AIwork #FutureOfWork

[Christopher Crawford @_CLCrawford]

Cyber resilience in Transit isn't optional anymore transit is essential . IEC 63452 helps guides operators and OEMs build to embed security into in every layer. Breakdown + details Details inside. ✅

#railcybersecurity #iec63452

https://t.co/TqHmxGZl87

[Bill Ackman @BillAckman]

The United States is about to become a vastly superior place to do business. We are already the safest and most militarily capable country in the world making us an extremely attractive place to live and raise a family in an increasingly dangerous world.

These facts have not gone unnoticed by successful people around the world who are not happy with what their own countries have become, and how their governments have turned against their hardest working and most successful citizens.

The problem with our immigration system is that it takes years for the talented, hard-working, successful, law-abiding immigrants to get in, and it takes only days for illegal migrants to cross the border and enter the country. We need to reverse this absurdity.

Imagine if we could vet applicants who wanted to bring their talents and resources here in a 45-60 day process as long as they met certain standards for excellence and character. This is not a particularly challenging problem and that’s before comparing it with catching a Super Heavy rocket with chopsticks.

We could open the floodgates for entrepreneurs, scientists, engineers, makers, designers and more who could replace the illegal migrants and those who have been consuming resources and bringing crime to our cities and towns.

The new immigrants would be force multipliers, bringing and building businesses which will drive our economy and employment. Consider the impact of deporting one criminal migrant and replacing him with a technology engineer and then do the multiplication.

The positive impact on our economy and growth rate would be extraordinary. Think Israel after its Russian migration when the best and brightest scientists and engineers left oppression and helped to build start up nation.

Think the United States in the 1930s when German scientists came and enabled our rocket, missile and atomic capabilities that helped win the war and build industries thereafter.

The time has come.

[Tier1 Alpha @t1alpha]

This shouldn't come as a surprise, but intraday volatility was through the roof today.

@profplum99 nailed it in this morning's note:

"And unfortunately for bears, elevated implied volatility that fails to materialize is often the catalyst for a rally."