Do you sometimes use underhype tactics for your videos?     See Less

seems all these cutting edge models are like early primordial soup, each vying for the suns rays, eventually some will merge into something more than the sum of their parts, in the mean-time     See More obvious and incremental updates that seem to happen, more or less, in sync with each other.    See Less

China will see the release and redesign computer cost/time for 1/100th circa 2 days.     See Less

15:39 so true. LLM is not always helping you for vibe coding.

I end up going back to use copy and paste of the LLM     See More code, instead of letting Cursor update the code for me.

When the code base become larger, vibe coding is slowing me down.    See Less

This guy didn't Bother to give us the Grok4 result on simple bench ... last time it was , NO API , now what's the excuse?????     See Less

The METR report needs it's own deep dive, because there are a bunch of nuances to the study like a mixture of folks with lots of AI tools experience mixed with people with almost no expe     See More milarly, there are many ways to code with AI from Vibe coding to auto-complete (but I'm writing most of it). I can tell you I'm way more productive when I let it eliminate stack-overflow and documentation lookups and execute my architectural ideas.    See Less

“ The Skynet funding bill is passed. The system goes on‑line August 4th, 2027. Human decisions are removed from strategic defense. Skynet begins to learn at a geometric rate. It becomes     See More re at 2:14 a.m. Eastern time, August 29th. In a panic, they try to pull the plug.”    See Less

Good job reading a twitter thread and dragging it out over 20 minutes! This type of content is worthless engagement slop     See Less

sounds like OpenAI did it🎉🎉
AGI
Definition:
–a model that beats all benchmarks, then without any training beats any new benchmark we make     See Less

AI is probably the only area where you have significant over hype and under hype at the same time often for the same thing...
I wonder what the impacts might be if "anyone" (cap     See More e 2K per month tier xD) has access to something that can solve IMO or similar stuff...
I wonder if the hallucination increase comes from the tool usage. Maybe the AIs get overwhelmed still by all these options. If that is true then it should go down if the underlying model between the tools improves. (One case on my private Tetris bench was that o3 just ignored what it did before and tried desperately to figure out what happened by using dozens of tools for minutes)    See Less