worktrees are good but they solve a different problem. Question is, if you have a lot of agent config specific to your work on a project where do you put it? I'm coming around to the idea that checked in causes enough problems it's worth the pain to put it somewhere else.
## Task Management
- Use the projects directory for tracking state
- For code review tasks, do not create a new project
- Within the `open` subdirectory, make a new folder for your project
- Record the status of your work and any remaining work items in a `STATUS.md` file
- Record any important information to remember in `NOTES.md`
- Include links to MRs in NOTES.md.
- Make a `worktrees` subdirectory within your project. When modifying a repo, use a `git worktree` within your project's folder. Skip worktrees for read-only tasks
- Once a project is completed, you may delete all worktrees along with the worktrees subdirectory, and move the project folder to `completed` under a quarter-based time hierarchy, e.g. `completed/YYYY-Qn/project-name`.
More stuff, but that's the basics of folder management, though I haven't hooked it up to our CI to deal with MRs etc, and have never told it that a project is done, so haven't ironed out whether that part of the workflow works well. But it does a good job of taking notes, using project-based state directories for planning, etc. Usually it obeys the worktree thing, but sometimes it forgets after compaction.
I'm dumb with this stuff, but what I've done is set up a folder structure:
And then in dev/AGENTS.md, I say to look at ai-workflows/AGENTS.md, and that's our team sharable instructions (e.g. everything I had above), skills, etc. Then I run it from `dev` so it has access to all repos at once and can make worktrees as needed without asking. In theory, we all should push our project notes so it can have a history of what changed when, etc. In practice, I also haven't been pushing my project directories because they have a lot of experimentation that might just end up as noise.
worktrees are a bunch of extra effort. if your code's well segregated, and you have the right config, you can run multiple agents in the same copy of the repo at the same time, so long as they're working on sufficiently different tasks.
I do this sometimes - let Claude Code implement three or four features or fixes at the same time on the same repository directory, no worktrees. Each session knows which files it created, so when you ask CC to commit the changes it made in this session, it can differentiate them. Sometimes it will think the other changes are temporary artifacts or results of an experiment and try to clear them (especially when your CLAUDE.md contains an instruction to make it clean after itself), so you need to watch out for that. If multiple features touch the same file and different hunks belong to different commits, that's where I step in and manually coordinate.
I'm insane and run sessions in parallel. Claude.md has Claude committing to git just the changes that session made, which lets me pull each sessions changes into their own separate branch for review without too much trouble.
I think this is for a different domain than software engineering, but it describes the basics, mostly difficult to get consensus on detailed requirements, often due to "personal" or "group political" reasons.
The skins market became a de facto stock market with extremely low volume (and therefore more manipulation). Someone eventually lost money, not necessarily the first link in the chain.
I once did a course with a paramedic on basic aid. We were discussing choking, which is a condition that really needs a 2nd party to intervene. Someone asked what to do if you live alone (with no close neighbours) - the answer was essentially ‘good luck’
The idea of the heimlich is to put sudden force on the diaphragm and force air upward. You can do that alone by pushing your upper abdomen against a chair back, counter, railing, whatever. Not something I've ever tried, but good to know about in case.
I know how to in theory, and I think I'm probably "above average" at calm-in-crisis, but my confidence that I'd calmly rescue myself via self-heimlich while unable to breathe is not high.
There is/was an AI Dungeon app running in the early/pre-ChatGPT days using GPT 3-ish, I think? Long term context was a real problem - story arcs were very.... drifty. A more modern agentic approach might help with this, doing multiple passes over the work to achieve consistency.
Back in the mid-to-late-1990s, when Lynx was the browser of choice, I encountered a collaborative online CYOA just like this. I have always thought it was called "The Neverending Story," although of course that's also the name of a movie. ...This person [3] also thinks it was called "Never Ending Story," and that it was still online as late as 2011(!).
You start at the "entrance" paragraph, where there are four or five choices — or, if you don't like any of those, you can just type in your own choice. The game then prompts you for what happens when someone selects that choice; and choice and consequence both go into some database on the server end, ready to be served to the next player. Anyway, you can follow the existing paths until you get to a dead end, at which point the game tells you to create at least two more choices and responses, so the next player will get to play a little longer.
As Gwern writes:
> So [any] player can ‘author’ an adventure by carefully curating a premise and then choosing actions and backing up and editing, creating a full-fledged scenario [...]
And that's exactly how it felt: By going down different paths at the start, you could navigate into a "Lovecraftian horror" subtree, an "alien abduction" subtree, a "romance" subtree, etc.
The technology at play here is just a minor extrapolation of the BASIC era's "Guess An Animal" game [1].
Plugging an LLM into the thing seems... well, frankly, it seems unnecessary. The core engine/database doesn't need an LLM for anything; the only thing the LLM would help with is coming up with new choices and new response paragraphs "less tediously" than we could do it in the 1990s. But the uncharitable way to describe that is: you could use an LLM to fill your CYOA game with AI slop, instead of hand-crafted texts that are meaningful to some real person in the world. ...Well, OK, maybe an LLM could provide a first rough pass on content moderation; or power a diagnostic like "Your new choice seems similar to this existing choice: [X] Are you sure you want to add this branch?". So there's places for LLMs in this. But I wouldn't use an LLM for content.
Several years ago I finally decided to try reimplementing the-thing-I-recall-being-named-"The-Neverending-Story" myself [2]; but I didn't get far, because (A) I've been too lazy to do anything requiring server-side hosting since Heroku went belly-up, and (B) opening such a thing up to the public means you're getting into the content-moderation business (Gwern also alludes to this) and ain't nobody got time for that. (See also "Why do you require an email address?" in [1].)
The guy in [3] doesn't think that [4] is the same website; but that's exactly the sort of locked-down, account-required, highly walled interface I'd have expected it to evolve into over the past 20 years. (However, it seems to have been essentially walled since <=2004, according to the Wayback Machine: [5].)
You’re right that really it’s collaborative production that’s the heart, and there’s no real benefit to LLMs.
It does feel like some classical explore/exploit algorithms could be interesting, but I imagine the challenge to any effort is really just getting enough high quality contributors
There’s a now quite dated comment from Merlin Mann: "Joining a Facebook group about productivity is like buying a chair about jogging.”
It’s fuzzy - but my recollection was Mann was a fairly renown productivity influencer (although I guess we wouldn’t have called it that then), who had an apostasy about it all.
I think he had a blog called LifeHacker and/or 36 folders (I don't know if they were his or just a writer, but I remember following him back in the day).
It's wild to think that was almost 20 years ago when Getting Things Done was going through tech circles as the organization method dujour. (I also equate the same time period with learning Ruby on Rails.)
Merlin ran 43folders.com[1]. That domain name was a reference to GTD's tickler files[2].
Merlin did not have anything to do with Lifehacker.com. Gina Trapani[4] founded Lifehacker[3]. But, Lifehacker often covered and was inspired by Merlin's work.
F1's official site has improved significantly since 2021, now scoring 75+ on mobile PageSpeed Insights vs ~45 back then. They've reduced initial payload by ~40% and implemented better image optimization techniques, though they still lag behind some team sites like McLaren's.
> Home runs, walks, and strikeouts now dominate baseball, with 35% of plate appearances ending without involving seven defensive players. This has reduced balls in play by 20% since 1980, creating longer games with less action
Do baseball fans ever discuss potentially changing the rules or game setup to mitigate this?
> Nearly a quarter of the way into the 2023 regular season, the rule changes MLB implemented this year are continuing to have their intended effects -- namely, a quicker pace resulting in a significant reduction in overall time of games, more hits and more steal attempts.
Yes. In the past MLB lowered the pitcher’s mound and widened the strike zone to benefit hitters, and there’s been talk in recent years of doing it again, or moving back the mound. The league also recently instituted the pitch clock, which helped hitters a little.
reply