Nico Neumann
Curious about opportunity. Building things. Documenting the mess.
I work on business systems, automation and software products with AI-assisted development. My current work includes GMT and personal agent systems. Read what I'm building now, October 2026.
The journal records an earlier stretch of May and remains here as an archive. Longer posts explain the systems, decisions and lessons behind the work.
Everything is plain text first. The repo is public. The analytics events are listed on /privacy. If anything you read lands or fails to land, the /feedback link comes straight to my inbox.
Journal archive
all entries →-
Day 16: tomorrow on the record
Preview entry, written the night before. Five targets on the record before the day starts, the first hour committed to outreach instead of build, a hard client deadline at end of day, Handyman moving in parallel on legal and Slice 2 UI, and jiu-jitsu late. Wins, losses, and lessons land when the day closes.
-
Day 15: moved the north star, did not move the message
Sprint day two. The plan still did not send. The bigger move was upstream, OneProposal stepped off the north star and the Handyman build, now Lydar, stepped on. A LinkedIn rebuild that ran four hours instead of two. A same-day brand pivot from Topa to Lydar after a gut-check, and the rename batch shipped clean. The wall-stare reset held for a second day, three hours of locked-in focus after. Eleven hours and forty-eight minutes on the clock.
-
Day 14: shipped the plan, did not send a message
Sprint day one in name, infrastructure day in practice. The math-to-ten post and the trilingual backfill went live, the blog's translation backlog closed, and not a single outreach message went out the door. An hour and a half on LinkedIn building the profile that has to send. Jiu-jitsu in the evening. And a focus reset that genuinely worked, fifteen to twenty minutes staring at a wall.
-
Day 13: hardened the machine, then drew up the sprint to sell
From morning jiu-jitsu through to the small hours, and by my own clock all of it is still the 26th. The morning went to hardening the system and clearing the receipts I had been dodging. The afternoon turned outward: an honest read of the OneProposal funnel, then a five-day sprint drawn up to land the first ten paying customers by Sunday. A full-day fast did something I did not expect to my focus. And the tracker I thought was off had been logging the whole time. Almost eleven hours of it.
-
Day 12: wrote the systems essay, then proved I still need one
Took the blog trilingual on Projects and Feedback, published the 'You fall to the level of your systems' essay, and caught the journal up. Then gave most of a nine hour tracked day to the screen. YouTube logged 4h 50m, about 2h 14m of it actually watched and the rest running while I worked. The day argued for systems and then showed exactly why I need them.
-
Day 11: a day with the family, and a real slice shipped
The day belonged to a family barbecue, from the morning shop to near midnight, and I am at peace with that. In the windows I had at the desk, the Handyman auth slice landed and got verified. The honest other half was an hour and a half lost to Instagram and Tinder.
Recent posts
all posts →-
I bet my next car on an AI system.
The tools, mistakes and decisions behind a real ChatGPT-assisted car purchase.
-
What I'm building now
An October update on GMT, AI-assisted delivery, personal agent systems, Lydar and OneProposal.
-
The math to ten
How the plan to reach the first ten paying OneProposal customers was reasoned: the unit math, the honest baseline, the channels the price forces, and the bet the week is built around.