Myles hamilton

Governance-first AI agent platform. One command install. Everything visible. Everything toggleable. Open source. Patent pending.

[Myles hamilton @ProjectLancelot]

I built an entire governed autonomous system in 18 days using Claude Code. Zero hand-written code.
Not a wrapper. Not a chatbot. A full enterprise-grade AI agent platform with:
→ Constitutional governance (Soul documents)
→ Risk-tiered execution (T0-T3)
→ Progressive trust graduation
→ Kill switches per subsystem
→ Encrypted credential vault
→ Immutable audit receipts
It’s open source and patent pending. 🧵👇

[Myles hamilton @ProjectLancelot]

@thsottiaux @argofowl instacart my friend you can get both. lmao. hell you can probably get your self a desk blender

[Myles hamilton @ProjectLancelot]

@thsottiaux what exactly do you mean?

[Myles hamilton @ProjectLancelot]

@dr_joel01 Why just why would studios or minis be a better use case for this or better yet how about actual servers. I would think any of these options would be more cost effective. But what do I know.

[Myles hamilton @ProjectLancelot]

@steipete @truebound @sama you always have the best and calmest replies for the haters! I Love it. keep up the good work. maybe some day people will get manners.

[Myles hamilton @ProjectLancelot]

@svpino i agree but the more interesting part is the engineer hiring isn't the same trigger mechanism. its not about hiring people to code its about hiring management and judgement.

[Myles hamilton @ProjectLancelot]

@thsottiaux I am jealous I would kill to be there. ( I'm kidding of course) but still I really really want to be able to go to these events.

[Myles hamilton @ProjectLancelot]

@ziwenxu_ @thsottiaux The Man is committed to the advancement of AI and making it work to better society and quite frankly hes making serious Dough. This is a choice he is sacrificing time now so he can enjoy time later. I commend him for his commitment.

[Myles hamilton @ProjectLancelot]

@AlexFinn I couldnt agree more and since you cant stop China from not releasing models that are uncensored you should just go full tilt and get models out.

[Myles hamilton @ProjectLancelot]

Google what are you doing ! Behind the frontier much?

[Myles hamilton @ProjectLancelot]

@the_codewala Windows 98 all went downhill from there

[Myles hamilton @ProjectLancelot]

i will also say that I think we are entering a time when model routers are more important then ever. you have to use the right model for the job to manage cost. This is what happens when you use the best model for everything. @thsottiaux I think you should add a native model router. maybe make it configurable so that we can chose what tasks to use what models for instead of having to remember to mange it manually.

[Myles hamilton @ProjectLancelot]

yeah I had never run into my weekly limits before and today i hit one for the first time. two days before reset. I do consistent work so definitely something changed. cant get too mad about it. I will say a reset for 80 bucks seems a bit high to me unless the reset is ontop of your normal reset usage. I cant justify paying 80 bucks for two days of usage only to have it reset at my normal interval and waist what ever is left of what I just purchased I would be open to it if the extra 80 dollar reset was just added to my normal usage when it resets.

[Myles hamilton @ProjectLancelot]

@Airaasayss 16 holes

[Myles hamilton @ProjectLancelot]

@thsottiaux @maria_rcks Morning Tibo! thanks for the reset! I will say I really liked the banked resets however. but beggars cant be choosers !

[Myles hamilton @ProjectLancelot]

@nisxant @thsottiaux haven't had a reset in what feels like forever!

[Myles hamilton @ProjectLancelot]

@Marsweep @C_3C_3 @Grok please compare the water usage of California's Golf courses cumulatively to the projected TerraFab facilties usage. please include information of ground water runoff from these Golf Courses and what levels of pollution they are contributing.

[Myles hamilton @ProjectLancelot]

@THECOUNTnews @Breaking911 @grok

[Myles hamilton @ProjectLancelot]

@amarifields_ Building https://t.co/RzdA9Herjm governance first agent OS. AI agents reason without consequence. We add consequence back in. Governance is architectural, not aspirational.

[Myles hamilton @ProjectLancelot]

@yashhq_22 The old saying if you build they will come. Yeah not so much !

[Myles hamilton @ProjectLancelot]

I built Lancelot with 5.5.

It’s a governance-native runtime for AI agents: policy engine, approval gates, kill switches, audit receipts, memory, model routing, HIVE sub-agents, War Room, and governed desktop control through UAB.

The token budget was ridiculous because 5.5 was reasoning across the repo, architecture docs, screenshots, execution traces, threat model, UX, and long design debates in one loop.

Earlier models could help build pieces. 5.5 helped keep the whole system coherent.

Current release-readiness pass: 7,216 tests passing, 90.5% Python coverage.

https://t.co/Yis0dOiaah