girvo

18 ms·

1. 421.▲
   by girvo 7mo ago

   I kind of want to use this to turn Wordgrinder into a Mac app haha
2. 422.▲
   by girvo 7mo ago

   > Although, surprisingly, built on top of absolutely incredible silicon. To me that's because thats a capital E "Engineering" driven task that Product can't get their grubby little mitts on and ruin.
3. 423.▲
   by girvo 7mo ago

   I'm currently experimenting with (trying to) fine tune Qwen3.5 to make it better at a given language (Nim in this case); but I am quite bad at this, and honestly am unsure if it's even really fully feasible at the scale I have acc
4. 424.▲
   by girvo 7mo ago

   > I’m sure we’ll all learn a lot from these early days of agentic coding. So far what I am learning (from watching all of this) is that our constant claims that quality and security matter seem to not be true on average. Depressingly.
5. 425.▲
   by girvo 7mo ago

   I had to setup a hook with terminal-notifier to get it to work myself
6. 426.▲
   by girvo 7mo ago

   That’s only if you use it in earnest, I would think
7. 427.▲
   by girvo 7mo ago

   That’s depressing. For 80% of my coworkers their LLM code is horrible. Only the seniors seem to use it well and not just spit out garbage
8. 428.▲
   by girvo 7mo ago

   Cockpit is great! My NAS (built on a weird “N17” AMD 7840HS laptop processor put into a desktop “server”mITX motherboard by those wizards in China) stuck in a Jonsbo N2 with 5x4TB Samsung 870 evos in ZFS raidz1 is entirely managed by it I k
9. 429.▲
   by girvo 7mo ago

   I pay for Apple One and yet the apple news app on my phone is still riddled with ads with weird AI generated people and horrible articles from crappy publishers pushing some other sensationalist garbage.
10. 430.▲
    by girvo 7mo ago

    That is working as intended. Google wants to kill side loading.
11. 431.▲
    by girvo 7mo ago

    Except for, yknow, the DoD and killer robots and mass surveillance drama.
12. 432.▲
    by girvo 7mo ago

    Harness :)
13. 433.▲
    by girvo 7mo ago

    GPT 5.4 keeps erroring out when I’ve tried it on long complex tasks. Might be works harness I guess.
14. 434.▲
    by girvo 7mo ago

    It’s the same in Australia. And they will live for another few decades most likely, so this only gets worse as far as I can tell. While the government removes all the benefits boomers and Gen X got to use to build their wealth, ensuring the
15. 435.▲
    by girvo 7mo ago

    > I think this passes the sniff test only if you're not too familiar with this neighborhood of the training set Which is totally fair, I'm honestly not! I haven't read much of that myself
16. 436.▲
    by girvo 7mo ago

    I will say this is one of the few pieces of prose I've read that was AI generated that didn't immediately jump out as it (a couple of inconsistencies eventually grabbed me enough to come to the comments and see your post details w
17. 437.▲
    by girvo 7mo ago

    Qwen3.5 (-plus, which isn’t actually open to be fair) is surprisingly decent I’ve found.
18. 438.▲
    by girvo 7mo ago

    I know it’s a bit of a tangent but man you’re right re. Gemini CLI. It’s woefully bad, barely works. Maybe because I was a “free” user trying it out at the time, but it was such a bad experience it turned me off subscribing to whatever thei
19. 439.▲
    by girvo 7mo ago

    Airbus' Ghost Bat equivalent?
20. 440.▲
    by girvo 7mo ago

    It’s not just me then! Hah good to know. It’s why I’ve started ignoring plan modes in most agent harnesses, and managing it myself through prompting and keeping it in the code base (but not committed)
21. 441.▲
    by girvo 7mo ago

    My annoyance with plan mode is where it sticks the .md file, kind of hides it away which makes it annoying to clear context and start up a new phase from the PLAN file. But that might just be a skill issue on my end
22. 442.▲
    by girvo 7mo ago

    I find gemini does that normally, personally. Noticeably worse in my usage than either Claude or Codex.
23. 443.▲
    by girvo 7mo ago

    That's fascinating: that is identical to the workflow I've landed on myself.
24. 444.▲
    by girvo 7mo ago

    Yeah that's the one. I've not managed to get close to the limits that the cheapest plan has. Though I did get to sign up at $3 a month which has been neat, too, seems that's gone now
25. 445.▲
    by girvo 7mo ago

    I'm using their $3 USD (currently, it will go up in price later I believe - edit: just checked and yeah, so the $10 one) lite plan, and I'm yet to get close to hitting the request limits when I swap to it once I'm out of Clau
26. 446.▲
    by girvo 7mo ago

    At work I regularly hit my 7.5mil tokens per hour limit one of our tools has, and have to switch model of tool, and I’m not even really a remotely heavy user. I think people don’t realise how many tokens get burned with CoT and tool calls t
27. 447.▲
    by girvo 7mo ago

    IMO you’re better off using qwen3.5-plus through the model studio coding plan, but ymmv
28. 448.▲
    by girvo 7mo ago

    I run Qwen3.5-plus through Alibaba’s coding plan (Model Studio): incredibly cheap, pretty fast, and decent. I can’t compare it to the highest released weight one though.
29. 449.▲
    by girvo 7mo ago

    To be fair to the model, I really can act like one sometimes.
30. 450.▲
    by girvo 7mo ago

    Semi-retroactively: my agent has a command to /compact and its then that I pop the interface. It gets opened automatically if the context is full, too, and then I've gone back and fed some recorded sessions into it as well days la