letting a second Claude decide whether to give permission for commands, severely reducing the number of times i have to give Claude permission for commands #productivity

started making bigger goals and working backwards from there #productivity

giving claude a postgresql mcp ah why did I wait so long to do this #productivity

Made Notion feeds visible in my Apple Calendar now I can make todo’s act how I want #productivity

Will be timeboxing more types of work each day to tackle all angles of business. Building is too attractive to me #productivity

Claude Code has removed the token counter. Not it is harder to see if you're being throttled. It is harder to debug why the hell things are go so slow. Anyone else running into this today? #productivity

#productivity claude indefinitely blocking when a command is indefinitely blocking is dumb. it should obviously be able to handle subprocesses parallelly.

#productivity learning: to improve Claude Code autonomy, maximize MCP use, and minimize CLI use. With CLI it very often tries to combine multiple commands (e.g. adding "| echo Command output: &2" or similar), and whenever it does, your command whitelist no longer works.

#productivity installing a hook to notify me with a sound or macos notification when claude code needs my input. looking into using opus - not clear on when claude uses what model.

Lack of errors with clear causes in Swift compiler is slowing down development severely. It’s mostly type inference stuff, eg “type ambiguous”. Will move towards relying as little on type inferencer as possible, to code defensively in this respect. #productivity

Both iOS development and backend development simultaenously accelerating! There's still too moments that i have to "help" it for me to effectively have a third agent, but these moments are definitely decreasing. #productivity

iOS development is becoming almost entirely agent-based without the need for xcode! #productivity

#productivity will start a third worktree and claude code agent to complete a full fledged web version to serve android users. the first and second will be working on backend end ios, respectively. #happenlist

#productivity Claude was consuming way too many tokens with ios build commands. Ran into my short term token limit. So I told it to write in CLAUDE.md to pipe outputs of large commands to files "./claude-outputs/<some file>" and add that folder to gitignore.

100th day on my WIP streak! When I started this streak, at first coming back everyday was a chore, but now it's becoming a habit. In my 20s I was working 16 hours a day locked-in for startups that I didn't own. That has become harder to do. Also, I became rusty in my habits. The positive change has come late in the streak, in the last few days really, now I've started splitting my focus between result and process, rather than only result (product). In some sense it is that shift of focus from the machine to "the machine that builds the machine". It's crazy how big of a blind spot this has proven to be for me. Will report here on further improvements. Currently I have a super low but at least nonzero $8MRR. I'll set a deadline of increasing that MRR by a nonzero amount before my streak hits 130. #productivity

#productivity moving up the productivity ladder by adding a second worktree and having two claude codes working at the same time, one on api and one on ios frontend. it is going well. #happenlist .

#productivity nailed that shit today. claude code rocks and it has test driven already built in. wow. ok moving on to next productivity improvements tomorrow.

#productivity on yesterday's challenge: worked test-driven, but didn't use ai to build that because i was in a place where there was no internet/reception today: again, applying 1) test-driven an 2) everything done through AI. Installing claude code as I speak. I use jet brains so i'm excited about dropping the second IDE (cursor) that i use just for ai.

#productivity am in a big overhaul of the "algorithm" behind #happenlist. Lots of lose ends, make me think a lot, and underestimate AI coding's ability to tie it together successfully. Also having many lose ends makes me have to think long about what to do next. These things will make me slow in finishing it. To get faster 1. I will force myself to let primarily AI code for every remaining part. 2. will ensure whatever i build today is surrounded by a test so i can instantly locally execute it so I don't have to think about what to do next (just run and fix the next error).