@FlintCasey i was not specifically talking about gpt-5, i was referring to what tradeoff i prefer, and how different people may define "good" in different ways
fwiw, i am having a solid time using gpt-5 for coding, and has become my main model (used sonnet before this)
@the_judge1111@HarryStebbings good observation tbh, didn't think of it like that
i think i mentioned this is my other tweet, i still think he is doing good admirable work
this feels like a close approximation of reality
Announcing MCP•RL: teach your model how to use any MCP server automatically using reinforcement learning!
Just connect any MCP server, and your model will start playing with it and (using RL) "learn from experience" how to use its tools most effectively!
i think @cohere is a value stock in the ai hype world
they have solid core talent, good revenue, canada discount, and i think their value proposition is the same as @thinkymachines
i think they have been around long enough to understand ai application
@davefontenot wild metrics, and congrats to the companies
id lead with that, i think it'd attract th right kind of founders compared to valuation chasers
we all know how that ended for zirp cos
this is truly a bad metric to track, an anti signal
most companies will come out with suicide rounds
what's the average arr? or, what's the dpi (ik, too early to tell)
im sure hf0 is solid, but if that's the metric to track instead of users or other proxy, it's a bad sign imo
$82 million. That's the average valuation companies who went out to raise after our last demo day hit.
HF0 is a place where you'll set an impossible goal, and blow it away.
6 spots remain.
Deadline. Friday. Midnight.
@chrisgpt big man i am rl pilled relax i think swe is cooked is a valid take on a reasonable time horizon
i just think these kind of takes are what make rl pilled swe is cooked crowd look kooky
devoid of any nuance
not your fault though, this platform promotes lack of nuance
A lot of people aren't impressed but you have to keep in mind this is crazy progress. A 2% gain in 2 months.
If scaled linearly, Claude 5.3 would be at 150% by mid next year
A lot of people aren’t impressed but you have to keep in mind this is crazy progress. A 2% gain in 2 months
If scaled linearly, Claude 5 would be at 92.5%% end of next year.