Sharing this in hopes it saves other people the data loss I experienced:CC's default setting is a 30-day data retention: https://github.com/anthropics/claude-code/issues/59248If you want to keep your transcripts, you'll need to update `cleanupPeriodDays` in your `settings.json`.
Factfulness helped to show me that, despite our brain's bias, the world is actually consistently getting better.
- Went deep on "what types of guidance even are there? what does giving good guidance mean?"
- Sampled my existing Claude guidance (CLAUDE.md, skills, hooks, etc.) and broke their guidance into "atoms"
- Categorized them by clustering, the same way Big Five was generated
- Generated a new candidate
- Then used independent agents to compare it against my existing corpus assuming that the new one would be worse
Working with it felt like working with a supersmart entity capable of generating very plausible-sounding but not-necessarily-true statements. The outcome certainly felt like an alien artifact, like nothing I'd make myself.
Only time'll tell if it holds up, but it sure had some interesting ideas.
1. put your thumbs on your ears
2. rotate your hands so your index fingers are on the base of your skull, middle fingers just above
3. now put your index fingers on your middle fingers and "snap" them down on the muscle at the base of your skull some 10-15 times
4. if your tinnitus goes away or reduces, it's caused by muscle tension instead of nerves
This blew my mind when I first tried it, but looked into it and it makes total sense: we all work on computers all day, necks get fatigued, and the impact forces the muscles to contract until they force-release, alleviating the tension-caused tinnitus.
It pulled back Plan 9, and I was shocked: this is exactly what we need today, as I'm convinced we need to think about minimizing agent permissions the exact same way companies do. Plan 9 was just too early.
One thing directories solve: they're great grouping mechanisms. "All the Q3 stuff lives in this directory"
I bet we move towards a world where files are just UUIDs, then directory structures get created on demand, like tags.
Source: been watching both these guys closely, as I've been building my own agent factory focused on security + learning: https://github.com/mieubrisse/agenc
I think Gastown is truly special, but I wanted something more focused on learning as I think that's the real bottleneck. So I built AgenC to make it trivial to roll learnings back into your Claude.
Hey folks,After rounds of doctors were unable to explain some visual symptoms I'm having, I recently started treating my high myopia as I would debug a production system: collecting data, building data pipelines, doing analysis, trying to spot trends.I'd love to talk to people with deep experience about what I'm doing, but my software engineering networks don't overlap at all with medical...
- https://chromewebstore.google.com/detail/df-utube-distractio...
- https://chromewebstore.google.com/detail/clickbait-remover-f...
- https://chromewebstore.google.com/detail/blocktube/bbeaicapb...
No, this is full-on war for control of your mind. And the adversary spends millions to hire teams of the world's best psychologists and engineers to deploy technology that never sleeps with the sole purpose of grabbing and keeping your attention.
Once I realized this, I started treating doomscrolling and Youtube rabbit holes not as personal insufficiencies, but as systemic failures in my psychological defense system. I started installing my own tech to keep me safe, and I am much, much happier.
Predictably, companies like Google try to disable the defenses (e.g. with Manifest v3, which was a garbage excuse to disable many defensive extensions). And so the war goes.
I once took a timed test with a section that had me translating a string of symbols to letters using a cipher, response being multiple choice. If you read the string left to right, there were multiple answer options that started with the same sequence of letters (so ostensibly you had to translate the entire string).
But if you read the string right to left, there was often only one answer option that matched (the right one). So I got away with translating only the last ~4 symbols, regardless of how long the string was. I blew through the section, and surely scored high.
I always wondered: did they realize this? Or did it artificially inflate my results?
And looking at the highest-entropy section felt natural to me, but only because of countless hours as a software engineer where the highest-entropy bit is at the end (filepaths, certain IDs, etc).
Is it really accurate to say I'm "more intelligent" because I've seen that pattern a ton before, whereas someone who hasn't isn't? I suspect not.
I worked there for 7y, and can confirm Nabeel's post is very accurate.
The general public thinks Palantir's a bunch of moral-less folks grinning over godlike power & privacy violations but in actuality it's a mashup of smart Silicon Valley + military folks trying to make data pipelines & analysis work around the globe, often in achingly bureaucratic organizations.
1. The products are powerful but complex, and were designed by smart, technical people for smart, technical people. They lack some random-person usability.
2. The average client employee isn't that technical.
Source: worked there for 7 years (my views are my own, of course)
> But luckily, success indexes less on IQ and more on consistency. The willingness to doggedly show up every single day can take you to some really suprising and amazing places.
Only, I never had the courage to make the leap to building an actual product out of it. Huge congrats to you guys; going to take a look at Projection and see if I can finally retire my homegrown stuff!
The concrete value-adds of cmdk (what fzf doesn't do out of the box):
1. Smart previews. You have to write the fzf preview command to preview text files, directories, images, PDFs, etc. cmdk gives you this out of the box.
2. Choose the command for you. If you rely on the fzf Ctrl-T shortcut, you have to type the `cd`, the `vim`, the `open`, etc. I find that most of the time the command is predictable based on the filetype, so cmdk allows you to skip this. When cmdk doesn't have a command for the filetype though, I supplement with fzf's Ctrl-T.
3. Be smart about which files you're searching to be performant. E.g. very likely users aren't going to jump to files in the "~/Library/Application Support", or certain system directories under "/", so give users the option to go to the parent directory without crawling the subdirectories.