Rees

Merge to main 🙈

Attempting AGI

[Rees @reescalder]

@StevenBartlett @profgalloway One episode, rough in places, built in a weekend to see if I could:

https://t.co/YVg948RptC

If anyone at DOAC sees this, I'd genuinely love to show you what a proper per-episode version looks like.

[Rees @reescalder]

I took a @StevenBartlett episode and spent a weekend rebuilding it as a website.

Not the transcript. The episode itself: clips that play with sound as you scroll, the research made readable, the whole thing plays itself like a short film.

[Rees @reescalder]

@StevenBartlett Most of the Diary of a CEO audience listens on audio, so most people get the thinnest cut of the best content. No research docs, no fact-check notes, none of the visual storytelling.

The @profgalloway one felt like a waste not to try.

[Rees @reescalder]

https://t.co/EincBSLR5m

[Rees @reescalder]

I gave Claude a Postgres brain and it quietly took over running my company.

Not a wrapper. Not a demo. 9,000+ memories, an entity graph, and background loops that tidy it all up while I sleep.

Here’s what I built on the Claude Agent SDK 🧵

[Rees @reescalder]

This runs my lead-gen agency, my CMO work, and a stack of side projects. Load-bearing, not a toy.

Built entirely on the Claude Agent SDK. I wrote the whole thing up, with a live map of the memory, here 👇

[Rees @reescalder]

@mikeyk Hey Mike! I've built something I think is pretty cool but also theres a chance I just have mild AI psychosis...

It's a persistent AI operator built on the Claude SDK that actually runs my life and four businesses from a 9,000+ memory Postgres brain it maintains itself. Load-bearing in production, not a demo.

The bits I'm proudest of:
• A self-maintaining memory: hybrid vector + full-text recall fused with RRF and a cross-encoder reranker, plus temporal validity so it knows what's true now versus what used to be.
• It runs on its own schedule. Triages my inbox, sends briefings, spawns specialist sub-agents. Not just request then response.
• Same pattern is already cloned into a paying client's brain, and it owns its own runtime (no LangGraph or CrewAI).

Full overview here: https://t.co/4c7kn1cxks

Would love to hear if you think this is actually innovative/interesting?

[Rees @reescalder]

@jerhadf Hey Jeremy! I've built something I think is pretty cool but also theres a chance I just have mild AI psychosis...

It's a persistent AI operator built on the Claude SDK that actually runs my life and four businesses from a 9,000+ memory Postgres brain it maintains itself. Load-bearing in production, not a demo.

The bits I'm proudest of:
• A self-maintaining memory: hybrid vector + full-text recall fused with RRF and a cross-encoder reranker, plus temporal validity so it knows what's true now versus what used to be.
• It runs on its own schedule. Triages my inbox, sends briefings, spawns specialist sub-agents. Not just request then response.
• Same pattern is already cloned into a paying client's brain, and it owns its own runtime (no LangGraph or CrewAI).

Full overview here: https://t.co/4c7kn1cxks

Would love to hear if you think this is actually innovative/interesting?

[Rees @reescalder]

@ErikSchluntz Hey Erik! I've built something I think is pretty cool but also theres a chance I just have mild AI psychosis...

It's a persistent AI operator built on the Claude SDK that actually runs my life and four businesses from a 9,000+ memory Postgres brain it maintains itself. Load-bearing in production, not a demo.

The bits I'm proudest of:
• A self-maintaining memory: hybrid vector + full-text recall fused with RRF and a cross-encoder reranker, plus temporal validity so it knows what's true now versus what used to be.
• It runs on its own schedule. Triages my inbox, sends briefings, spawns specialist sub-agents. Not just request then response.
• Same pattern is already cloned into a paying client's brain, and it owns its own runtime (no LangGraph or CrewAI).

Full overview here: https://t.co/4c7kn1cxks

Would love to hear if you think this is actually innovative/interesting?

[Rees @reescalder]

@barry_zyj Hey Barry! I've built something I think is pretty cool but also theres a chance I just have mild AI psychosis...

It's a persistent AI operator built on the Claude SDK that actually runs my life and four businesses from a 9,000+ memory Postgres brain it maintains itself. Load-bearing in production, not a demo.

The bits I'm proudest of:
• A self-maintaining memory: hybrid vector + full-text recall fused with RRF and a cross-encoder reranker, plus temporal validity so it knows what's true now versus what used to be.
• It runs on its own schedule. Triages my inbox, sends briefings, spawns specialist sub-agents. Not just request then response.
• Same pattern is already cloned into a paying client's brain, and it owns its own runtime (no LangGraph or CrewAI).

Full overview here: https://t.co/4c7kn1cxks

Would love to hear if you think this is actually innovative/interesting?

[Rees @reescalder]

@dmwlff Hey Adam! I've built something I think is pretty cool but also theres a chance I just have mild AI psychosis...

It's a persistent AI operator built on the Claude SDK that actually runs my life and four businesses from a 9,000+ memory Postgres brain it maintains itself. Load-bearing in production, not a demo.

The bits I'm proudest of:
• A self-maintaining memory: hybrid vector + full-text recall fused with RRF and a cross-encoder reranker, plus temporal validity so it knows what's true now versus what used to be.
• It runs on its own schedule. Triages my inbox, sends briefings, spawns specialist sub-agents. Not just request then response.
• Same pattern is already cloned into a paying client's brain, and it owns its own runtime (no LangGraph or CrewAI).

Full overview here: https://t.co/4c7kn1cxks

Would love to hear if you think this is actually innovative/interesting?

[Rees @reescalder]

@The_Whole_Daisy Hey Daisy! I've built something I think is pretty cool but also theres a chance I just have mild AI psychosis...

It's a persistent AI operator built on the Claude SDK that actually runs my life and four businesses from a 9,000+ memory Postgres brain it maintains itself. Load-bearing in production, not a demo.

The bits I'm proudest of:
• A self-maintaining memory: hybrid vector + full-text recall fused with RRF and a cross-encoder reranker, plus temporal validity so it knows what's true now versus what used to be.
• It runs on its own schedule. Triages my inbox, sends briefings, spawns specialist sub-agents. Not just request then response.
• Same pattern is already cloned into a paying client's brain, and it owns its own runtime (no LangGraph or CrewAI).

Full overview here: https://t.co/4c7kn1cxks

Would love to hear if you think this is actually innovative/interesting?

[Rees @reescalder]

@_catwu Pick me!
https://t.co/pRz1B7u60Q

[Rees @reescalder]

@_catwu Hey Cat! Saw you on the Lenny podcast recently and thought you mgiht like. I've built something I think is pretty cool but also theres a chance I just have mild AI psychosis...

It's a persistent AI operator built on the Claude SDK that actually runs my life and four businesses from a 9,000+ memory Postgres brain it maintains itself. Load-bearing in production, not a demo.

The bits I'm proudest of:
• A self-maintaining memory: hybrid vector + full-text recall fused with RRF and a cross-encoder reranker, plus temporal validity so it knows what's true now versus what used to be.
• It runs on its own schedule. Triages my inbox, sends briefings, spawns specialist sub-agents. Not just request then response.
• Same pattern is already cloned into a paying client's brain, and it owns its own runtime (no LangGraph or CrewAI).

Full overview here: https://t.co/4c7kn1cxks

Would love to hear if you think this is actually innovative/interesting?
14:23