Writing / one human, one second brain
Notes from building a local, private second brain and running an engineering org that works with agents. Written for LinkedIn first, kept here so they last.
Estonia's weekly developer podcast, a Steam Machine instead of a basement GPU rig, and a live demo where the assistant said it had no data.
An hour on Estonia's weekly developer podcast about the Steam Machine that runs my second brain, with chapters, the measured numbers and the full transcript.
General Whisper turns Estonian into words that do not exist. TalTech's finetune does not. Measured on real recordings, on one CPU, in my living room.
A Steam Machine with 8 GB of VRAM, running sight, hearing, voice, imagination and memory for a second brain. One model at a time, no cloud, no API bills.
Most AI memory systems either forget everything or remember garbage. The fix is a holding inbox, a read-only rule, and one demotion: memory is a hint, never evidence.
Mine and a friend's, trading failures and fixes through one shared file. Nothing private in the room, and every real change still a human decision.
A local, private assistant built around one rule: never guess. It reads, it digs, it cites, and it works the night shift. I decide.
Half the users of our project tracker will never open its UI. Humans, agents and other systems are all users now, and the API surface is the product.