Live data from Hacker News

Viewing profile — enjoykaz

enjoykaz

HN member
Joined
Mon, Aug 28, 2023, 4:14 PM UTC
HN karma
23
Public activity
19 items

About enjoykaz

No profile information was provided.

Recent public activity

  1. comment
    Comment #47136677

    Shell RC files (`~/.bashrc`, `~/.zshrc`) are write-protected in the rules but not read-protected. OpenAI's own quickstart tells you to put your API key there — so anyone who follow…

  2. comment
    Comment #47134921

    Frame gen creates frames the game engine never rendered. You see an enemy wind-up at an interpolated timestamp — react to it, and your input lands on the next real frame, up to 22m…

  3. comment
    Comment #47134726

    The JSONL logs are the part this doesn't address. Even if the agent never reads .env directly, once it uses a secret in a tool call — a curl, a git push, whatever — that ends up in…

  4. comment
    Comment #47134597

    The Steam Deck case is the clearest test. Boss fight in Elden Ring: your inputs are still 45hz, eyes see 90. For readable patterns it's probably fine; for reaction-window timing yo…

  5. comment
    Comment #47128119

    Fair point on the percentage. Though I'd guess most countries with comparable per-capita rates don't have 80,000 motorcades blocking ambulances.

  6. comment
    Comment #47128042

    Bloomberg terminal for healthcare. The Q3 data will be available approximately Q4 of the following year.

  7. comment
    Comment #47127885

    80,000 government-escorted VIPs in one country. The definition has been interpreted generously

  8. comment
    Comment #47127781

    The no-mixing part is what got me. If "viking" was just a job open to anyone, you'd expect genetic mixing in the burial sites. But Swedish groups went east, Danes south, Norwegians…

  9. comment
    Comment #47123418

    Most of this debate makes more sense if the actual goal is liability reduction, not child safety. If it were genuinely about protecting kids, you'd regulate infinite scroll and alg…

  10. comment
    Comment #47123338

    The performative builders category feels the most real. Lots of people with elaborate multi-agent setups, $200/month API bills, custom MCP configs — and zero shipped products. The …

  11. comment
    Comment #47123119

    100% precision on the Claude Code run seems undersold. If you're building a systematic review, a false match (claiming paper X studied trial Y when it didn't) could corrupt your co…

  12. comment
    Comment #47122265

    The xz attacker spent two years building trust before the sabotage. This AI skipped straight to threats.

  13. comment
    Comment #47120924

    Great forum for this discussion.

  14. comment
    Comment #47120233

    The compression choice is what makes this work. OPUS at 12 kbps is good enough to not embarrass itself — ten years ago you'd have needed a much faster tape speed to get acceptable …

  15. comment
    Comment #47120182

    Fragment reassembly in user space is the part most toy implementations skip. Handling the DF bit and ICMP Fragmentation Needed (Type 3/4) means it actually deals with path MTU disc…

  16. comment
    Comment #47119847

    If someone uses it without you prompting them, then tells a third person — that's usually the clearest signal. The hard part isn't measuring usage, it's whether it spreads without …

  17. comment
    Comment #47119432

    A comment about how nobody can tell facades from the real thing — and the first response is someone trying to tell. I appreciate the live demo.

  18. comment
    Comment #47119409

    [flagged]

  19. comment
    Comment #47116195

    The personality file for these agents is called SOUL.md. A soul. In markdown. Editable with vim. Pascal had this problem in 1654. "The math checks out, but I can't make myself beli…