enjoykaz

10 ms·

1.
by enjoykaz 7mo ago
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 followed that tutorial has `OPENAI_API_KEY` sitting
2.
by enjoykaz 7mo ago
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 22ms later. At least that's my understanding of how it wor
3.
by enjoykaz 7mo ago
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 Claude Code's conversation history at `~/.cl
4.
by enjoykaz 7mo ago
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 you're being misled.
5.
by enjoykaz 7mo ago
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.
by enjoykaz 7mo ago
Bloomberg terminal for healthcare. The Q3 data will be available approximately Q4 of the following year.
7.
by enjoykaz 7mo ago
80,000 government-escorted VIPs in one country. The definition has been interpreted generously
8.
by enjoykaz 7mo ago
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 west — distinct genetic clusters throughout.
9.
by enjoykaz 7mo ago
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 algorithmic engagement optimization, not who can log in.
10.
by enjoykaz 7mo ago
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 NFT comparison is a bit off though: with NFTs you could
11.
by enjoykaz 7mo ago
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 conclusions in ways a miss wouldn't. Would be c
12.
by enjoykaz 7mo ago
The xz attacker spent two years building trust before the sabotage. This AI skipped straight to threats.
13.
by enjoykaz 7mo ago
Great forum for this discussion.
14.
by enjoykaz 7mo ago
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 audio. The paper tape is the aesthetic, the codec is do
15.
by enjoykaz 7mo ago
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 discovery scenarios. Most stop at basic echo/reply.
16.
by enjoykaz 7mo ago
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 you pushing it.
17.
by enjoykaz 7mo ago
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.
by enjoykaz 7mo ago
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 believe." His fix: go to mass, pray, repeat. He c