🥇 in 40+ division with takedown in overtime and leg sub in final 🦵 🥈 in 35+ wrestled same (smaller) guy for two matches and two overtimes, couldn’t take him down once Decided to stand and wrestle instead of guard pulling 🤼‍♂️ ADCC Orlando — @ana_analytics_ Trials tomorrow! 💪

These are 10x equally sized buckets. So the effect is even larger if you look at bottom 5% of news sentiment instead of bottom 10%. We aggregate the same news into one story. So there would be one NVIDIA news story here (per day) even if many people are writing about it.

According to our research, the best stocks to short (over 5-10 day period) come from moderately bad news, not from the most negative stories. The sentiment scores factor in mostly "how negative does this headline sound" plus more weight assigned to more sources.

Google / Meta TPU partnership was seen as bad news for NVIDIA yesterday. The stock was down -5% yesterday. The next day? NVIDIA back up, while Google is down a little bit. When we looked at news sentiment effect on stocks, you can see that positive news generally good, negative

Comms marked safe from AI for another day I asked multiple models how NVIDIA could've tweeted without seeming insecure about Google Their answer was to attack Google harder lol

Assuming this research is hard but ultimately pays off… Bearish for models getting commoditised any time soon. Bullish for NVIDIA and Google TPUs probably. Even if eventually models more capable at lesser size

The gains from bigger and more efficient models, trained on more data… are diminishing perhaps. But more importantly certain “common sense” capabilities we want from models — are not emerging naturally from more scale. Easy to see what labs are spending billions on researching.

The age of scaling may not be over. It is not the end, it is not the beginning of the end…. but perhaps it is the end of the beginning.

Great conversation. Starts with Ilya opining on why * AI capabilities seem far ahead of economic impact * success on hard metrics — for coding, etc — don’t seem to translate to realistic human level improvement on similar tasks Is the the RL? Or the training data

Google, but even more so Meta, have large internal use case for GPUs to make money -- optimizing Ads. Hence they have flexible and positive value uses for their build-outs even if they "over invest" a little. Unlike AWS or Microsoft, who rent out capacity to others.

Pretty incredible that Google more than doubled since low point in April, after being almost flat for four years. Much of this because of Gemini. Still lower market cap than NVIDIA.

Today's free issue of The Diff: the death of local news and of normal middle-class jobs in media is, unfortunately, a result of a more efficient labor market for an unusual job. https://t.co/vvALIt3V7m

After a HUGE week of Google, xAI and OpenAI releases, @Anthropic has now entered the Arena with Claude Opus 4.5! Claude Opus 4.1 currently holds a strong #4 on the WebDev leaderboard (powered by Code Arena) and ranks #7 in the super competitive Text Arena. How much stronger

Whatever you think of Azov guys, pretty sober assessment of the situation. Neither side will respect assurances made in any Russia / Ukraine agreement. The front line isn’t moving that much, but improving in pro-Russian direction.

Wonder if this plays out like Apple and android where the average person would never consider using Android no matter how cheap/good it is. I don’t think a single suburban mom or college kid would use Gemini over ChatGPT at this point, the customer behavior is so ingrained.

I said it was coming but guess a week early lol

Yep, this is pretty much what happened. There is no way to tell the algo that you want non-ragebait, and non-bait content in general. Perhaps this is inevitable in a text-based public posting system. Still lots of great content here but skew is toward bait...

This is from the best independent pro-Russian English language account. The tragedy -- apart from loss of life and years of war (this has to be said) -- is there is not that big a gap between what the Russian side would accept, and what pro-Ukraine but rational people will tell

FYI

The right attitude.

Interesting that people held on very dearly indeed to the Growth at an Unreasonable Price names. Yes, HOOD got wacked. FIG down 6%. PLTR down 5%. But SHOP not even down 1%. NET and AXON only down 2%. SNOW only down 3% All seems very orderly to me. No real fear yet