• 0 Posts
  • 175 Comments
Joined 3 years ago
cake
Cake day: June 30th, 2023

help-circle



  • Been on a GLP-1/GIP compounded injection for a year. Lost 90lbs.

    I physically have a hard time overeating on fats/protiens. Like I want to vomit. Meat consumption is way down from before.

    Simple carbs like starches and sugar I can overinduldge on but seem to get sick faster than before taking the medication.

    It mostly helps with food related cravings but its not a silver bullet. I can still fallback to emotional eating under high stress and fluctuate +5 lbs with water retention. Its just much easier to recover.

    I lost about 1/3 the weight on my own tracking everything. With the medication it tripled the rate of loss and I didn’t need to track anymore. I’m much more able to trust my body and hunger signals developing a healthier relationship with food.


  • strongly encouraged for me means my entire bonus is tied to skilling up with LLMs.

    “cite-or-stop” with journaling, role definitions, subagents to control context windows, and shifting as much of the work as possible to deterministic scripts has yieled pretty high-quality results.

    Expensive af though. Like the self-review skill I maintain before engineers put their code up for human review will find 30-40 things on average, where only a small handful are false positives, can burn 10% of a $20/mo plan’s utilization in a single run on a moderately sized PR. The default one my company setup in our source control usually finds 1-5 things and only 0-2 are of any value.









  • sloppy_diffuser@sh.itjust.workstoLinux@lemmy.ml*Permanently Deleted*
    link
    fedilink
    English
    arrow-up
    11
    arrow-down
    1
    ·
    25 days ago

    https://grapheneos.org/usage#web-browsing

    Chromium and their particular fork have much better exploit hardening via sandboxing.

    My understanding is Firefox has better anti-fingerprinting and uBlock origin via manifest v2 support (or v2 features ported to v3).

    The argument often used is malicious ads. Sandboxing and hardening largely mitigates ads that contain exploits, but it doesn’t protect against social engineering, crypto mining, tracking, etc.

    So I guess it comes down to your threat model and desired experience.

    I personally prefer the uBlock origin experience, but an ad free experience and escape from targeted advertising was my target opsec when venturing into privacy.




  • My experience is you have to close as many degrees of freedom as possible. Its tedious as hell for generating quality code.

    Its great at debugging if you require it to manage its context window by delegating tasks to scoped subagents, generate evidence with references, and verify that evidence with a minimal reproducible example. Expensive… I’ve seen them run for a solid 30 minutes before responding back (not including the “thinking” log), but it usually finds the issue.

    A similar technique can be used for code generation but again it burns tokens and takes awhile. Have it generate and verify isolated reference implementations for anything nontrivial. Much easier to review with the rest of your domain and layered on complexity stripped out. The “thinking” log is interesting to watch as it bangs it head against bad assumptions or documentation and needs to start digging into dependency source code to work it out.

    Only then apply the implementation to your project from the reference implementation. Takes breaking down the tasks though to small enough units and closing those degrees of freedom.

    Anecdote on degrees of freedom: This one didn’t require a reference implementation in particular. I was reviewing a PR (LLM assisted, I wasn’t the authoring dev) to add signature validation to OAuth tokens. It duplicated the entire header/token parsing logic. It needed that path closed with a pointer to where the existing logic was and explicit requirements to enhance it. Refactor was great upon reviewing and the PR size was reduced by more than half.





  • what did you like more about rclone than Cryptomator?

    I wanted to leave Dropbox and ran across it. I liked the number of supported backends under one tool. I use it to access things beyond Backblaze like gdrive, SharePoint, OneDrive, Proton Drive. Well documented config file format. I was able to manage the config with Nix due to this.

    Is it suitable for sync, or is it more for backups

    It works great for one way sync. Bisync I never got working well enough to trust it. Bisync is nice for 3-way merges (two devices modifying files on the same cloud drive). Dropbox, gdrive, OneDrive win here. I’ve learned to live without it.

    I’m ideally looking for near-ish to real-time sync for contacts, notes, files, and pictures

    On a computer the fuse mounted volumes are near live. Cahce locally in a VFS. Anything else you’d have to script probably. There is rclone-watch but can’t say I’ve tested it

    With Round Sync you can browse with live refresh when you move between directories, but syncing would be on a schedule. Looks like a 15m interval is the fastest frequency.

    Are there any frontends for Linux you’d recommend, or do you script out the functionality you’re looking to implement?

    I mostly just mount on login with the VFS cache. Use my normal file browser. One command per mount. Its rare (practically never) that I need to work on something without internet, so I don’t deal with trying to script syncs. I tried in the early days of playing with it, but fuse mounts ended up meeting my needs.

    No GUI that I use outside of my normal file browser. The only thing I need to use the CLI for is cleaning up soft deleted files and old versions (Backblaze specific thing).