Very odd.. if there’s such high demand for this things .. and more demand than supply .. it certainly is odd that we see an explosion in compute sellers on X…
😊 retweeted
I'm so glad AI killed LeetCode-style coding interviews.
I've talked to young AI talent, and most would simply skip companies that still use LeetCode interviews.
I agree with Felix. Taste, empathy, principled thinking, business sense, and agency are what really matter.
The coding interview is dead, at least for my team at Anthropic. I've seen enough to know that it no longer provides signal. I removed it from loops this month.
I have a spicy, optimistic take - I think there's an opportunity for junior engineers: felixrieseberg.com/hunger-be…
😊 retweeted
This is something I am writing about for some time.
We can have 120B tokens per living person on earth with DS 4.1 Flash 1017 quality models by end of 2027 in terms of generation capacity of 25 GW (250B rental) & Vera Rubin class servers.
We can also have 1.2B tokens (1200 M) of Kimi 3.1 Pro quality per living person by end of 2027. Every person - kids, old people, people in Congo, Papua New Guina and so on.....
Assuming there are only 1B serious users from 8B total, it is:
- 960B tokes per person of DS4.1 class
- 9.6B tokens per person of Kimi 3.1 class
Yes, there would lot of slack in the system. But, this is a lot considering we are leaving another 25GW for training and post training (SemiAnalysis estimate is 50GW coming online by end of 2027).
I don't think training needs nearly that much by end of 2027 considering Anthropic spend 15B till date on training compute.
WDYT?
This glut will however be very good for model labs.
xitter.cf/bookwormengr/status/21…
Too cheap to meter: It is 0.025 cents (not dollars) per million tokens.
Look at the insane performance numbers for Kimi 3.1 on Vera Rubin
- 90$ worth (45M tokens * 2$ per million for input, assuming most tokens are input) of Kimi 3.1 tokens for 1$ of TCO at 100 tokens/sec interactivity. Insane margins. Lot of scope for players to cut prices.
It is true: The more you buy, the more you save 🙇♀️
- With theoretically same uplift for DS v4.1 Flash on Vera Rubin from Grace Blackwell (1600M per dollar TCO), there will be nearly 4000M tokens per dollar of TCO.
It is 0.025 cents (not dollars but cents) per million tokens.
(Grace Blackwell chart below for comparision. Then look at SemiAnalysis chart).
Explain me again how there is going to be 250B of inference GPU rental revenue next end year (25GW of inference ompute), and even higher GPU revenue year on year to justify 1T USD of per annum investment that will go on in perpetuity.
It is nearly 30$ per living person of GPU compute (8 billion population) and 120B tokens per living person (babies, kids, old people included) at 0.00025$ generation cost for a million tokens of DS v4.1 quality.
That is the level of consumption we need.
Either we have a glut or everyone has AI waifu, BF, dad, mom whatever they desire. I don't see everyone having a personal robot before 2027. I don't know about you.
Let me know if you disagree. These are relatively idealized scenarios. There would be lot of slack in the system and generation may be few tens of billions per living person, but even then it is a lot!!!
😊 retweeted
🛰 Great news! Ukraine has launched its first satellite for its own communications network
Himera, together with Denmark's GomSpace, has launched the Obria Nano satellite into orbit.
With its help, Ukraine will be able to develop its own communications infrastructure, relying less on foreign satellite systems.
The first test satellite of Ukraine's future sovereign satellite communications network is already undergoing testing.
😊 retweeted
i just don't see where the demand for this is today
is there all this hidden deepseek usage that we're not seeing? whenever we talk to providers it's clear all of their other customers are 100x smaller
Gavin Baker Believes Demand Will Solve Neocloud Financing Issues
"An open-source model, DeepSeek, said that they are monetizing at $38 per GPU hour. Last time I checked, I hadn't heard of anybody renting GPUs anywhere near that high. So maybe this explains why Blackwell prices have been up 17 days in a row."
Jensen Huang: "Yeah."
$NBIS $IREN $CRWV
This is actually key to keep track of …
Dario Amodei: some tokens Claude produces are worth a few cents, others could be worth tens of millions of dollars.
His examples:
1. Telling someone to restart their Mac: the model has said it 10 million times, worth a dollar or less
2. Telling a pharma company to move an aromatic ring to the other end of a molecule: potentially tens of millions
3. Expect pay-for-results pricing, or labor-like pricing by the hour
4. It's a new industry, so many models will be tried
Today pricing is simpler: per the S-1, usage-based pricing was 80% of H1 2026 revenue and subscriptions 20%.
Full breakdown here:
theaiopportunities.com/p/ant…
😊 retweeted
BREAKING NEWS🚨: NVIDIA just submitted its first Rubin NVL72 Kimi K3 submission, and it can reach over 850 tok/s/user, even at 2.8 Trillion parameters, with over 3x higher throughput on @vllm_project!
Great work by the NVIDIA team, and excited for Rubin Software 1.0 GA coming soon!
They are scrambling for data
Discord is about to put your public messages on Google
Discord will start making messages from Discoverable public servers searchable on Google and other search engines from mid-December.
The feature is ON by default in most regions, but you can opt out:
User Settings → Data & Privacy → turn off “Show my message content from Discoverable channels on web pages.”
In the EU, UK, Canada, Brazil, Australia and South Korea, it will be OFF by default.
Private servers, DMs and group DMs are not included. Usernames and avatars won’t appear alongside indexed messages.
