A stack use-after-free sitting in every Linux distro for 15 years is the story that should scare fintech engineers more than any AI headline this week. We audit dependencies for licensing and CVEs on a schedule, then run the same base kernel unpatched for years because "it's Linux, it's fine."
https://t.co/W0J3EWAokv
@GergelyOrosz The scarry part is how come the "upload your files to our server" can be the default option? (not to mention why they need this in the first place)
I don't buy this run /privacy command, and everything will be good narrative.
This is a reasonable ask, but the labs are probably being vague because they genuinely don’t have stable projections.
We’re still in the ‘subsidize aggressively to win the platform’ phase. The moment competition intensifies and inference costs stop falling as fast, pricing pressure will hit
the gap is where most founders quietly lose.
they hear the insight, feel the spark… then spend days ‘refining’ it until the moment passes.
the ones who win just ship by afternoon.
not because they’re reckless, but because they understand that speed is how you buy more shots at getting it right
Genuine question for anyone running Claude Code and OpenCode side by side in prod: does that 33k vs 7k token overhead actually move your latency and bill, or does prompt caching make it a non-issue once you're at scale? Curious what the numbers look like outside a benchmark post.
@emollick It's quite mental. We’re no longer just describing profound ideas; we’re instantiating them. The barrier between “I understand this concept” and “I can now explore/play/build with it” has collapsed. That’s the part that feels genuinely new in human history.
@coreyhainesco Love this. Funny that nowadays most of the sites uses the same pricing layout.
Curious to see if you find one that differs but still converting good. I'll watch that space!