matt

building a second brain

[matt @matttli]

published my readwise wiki on my personal site: https://t.co/3AgTDKY97W

it's all AI stuff i'm trying to absorb and learn from.

still figuring out the best way to actually learn from it, whether that's reading through pages, querying it via my second brain agent, or a speech-to-speech conversation via pipecat. open to ideas.

[matt @matttli]

https://t.co/qlNJy1R6M9

[matt @matttli]

@lexfridman You do matter. Your feelings are relatable and comforting, especially coming from someone I respect and admire. Haters gonna hate, but your authenticity is a reminder to keep doing what we do with honesty and openness.

[matt @matttli]

Been running and learning from daily AI briefings with NanoClaw, but also drowning in things I may never get to reading, watching, or listening to...

Inspired by @karpathy's wiki pattern, I'm experimenting with feeding my saves from @readwise to an agent that compiles the content into a wiki, and my LLMs are instructed to reference it while I work. Knowledge compounds without me consuming everything.

First run: 65 Readwise saves → 12 interlinked wiki pages. Tempting to add everything I can to it, but I like the idea of having it personally curated with what originally drew my interest and curiosity.

#buildinpublic

[matt @matttli]

Would it really be that hard to teach LLMs what day and time it is?

[matt @matttli]

@larsencc I’ve thought this too. Tech folks and enthusiasts do, but it definitely requires time and effort to understand and keep up. It makes me believe that those not in the tech world have no idea what’s coming

[matt @matttli]

@walkojas @philkellr You mean they’ll be crashing your car and punching you in the chest? 😂

[matt @matttli]

I built a tool that tracks every change to your CLAUDE.md and AI instruction files. Local dashboard with diffs, timeline, and file browser. Free, open source, all data stays on your machine.

npm install -g @mattli/dotmd

https://t.co/L8vYKOu7zo

[matt @matttli]

@nbaschez My PM mind automatically goes to smaller scope, and planning for iterative development. Is that absurd in this vibe coding era we are in?

[matt @matttli]

@Voxyz_ai thanks for this. i've been in the midst of comparing and struggling to commit to one or multiple. i generally hate the idea of using more tools than needed, but great to see that they can be complimentary

[matt @matttli]

Experimented with scraping 'drop your project' threads on X and categorized them by problem space. The most interesting part was seeing where builders are clustering vs. where nobody's building at all... #buildinpublic

[matt @matttli]

@Papito_192 chillin. just trying to keep up with with all the ai news/tools

[matt @matttli]

@AdamFard_ just messing around with my MacMini/Nanoclaw system for research, planning, and development. cheers

[matt @matttli]

@vineerpasam Perhaps our agents will need it https://t.co/fNqnFNGEyL

[matt @matttli]

@rohitdotmittal I just got back on X recently. It feels like absolute chaos. I’m learning a crazy amount being immersed in it, but being really mindful of burnout and overwhelm

[matt @matttli]

I find myself regularly trying to understand what exactly it is I’m building, but the truth is… I don’t really know. A system to help me research, validate, and build, sounds great, but the process is full of constant experimentation and fine tuning. What continues to amaze me is how this custom built system is tailored to how I work, and both casually and profoundly changing the way I manage my life. #buildinpublic #nanoclaw

[matt @matttli]

At the moment, I’m using NanoClaw to schedule research and analysis tasks, and have also naturally began using it for note taking and a todo list. Have so many repos, frameworks, plugins to investigate. Looking forward to building a full development workflow though and continuing to see what’s possible

[matt @matttli]

i think i just realized how fucking immersed i am in this ai shit right now
#buildinpublic

[matt @matttli]

Lean principles apply to building AI systems too. I've been spending a lot of time reverse engineering research tasks. Seems like a recurring theme with AI builders is quality control and observability
#buildinpublic