Exabite AI (@ExabiteAI) header
Exabite AI (@ExabiteAI) profile picture

Exabite AI

@ExabiteAI

Data so good you want to eat it. Account mostly ran by alana. Email me! [email protected]

 Your hard drive
exabite.ai
Joined August 2025
  • Tweets41
  • Following153
  • Followers283
7Media
Photo by Exabite AI (@ExabiteAI)Photo by Exabite AI (@ExabiteAI)Photo by Exabite AI (@ExabiteAI)Photo by Exabite AI (@ExabiteAI)Photo by Exabite AI (@ExabiteAI)Photo by Exabite AI (@ExabiteAI)Photo by Exabite AI (@ExabiteAI)
come hunt tokens with us
near (@nearcyan) profile picturenear
@nearcyan24 Jun 2025
My job? I'm a rare token hunter. I track down dead languages in Tibetan monasteries, decrypt Tesla's private journals, chase whispers of pre-contact Amazonian dialects. The AIs pay top credit for tokens they've never tasted, you know. Work is work, even if it's for the machines.
Image from near (@nearcyan)
1
28
5,344
I ran the Pangram 4 model card through Pangram
Image from Exabite AI (@ExabiteAI)
4
145
What a wonderful thing! 🩷 Congrats to Magic
Magic (@magicailabs) profile pictureMagic
@magicailabsSep 8
Frontier pretraining is said to be a big-lab-only game. We don’t have 100k chips yet, so there’s only one way: algorithmic efficiency. Our new recipe matches DeepSeek V4 Pro’s pretrain using 50x less compute – that’s roughly half the FLOPs used for GPT3, or ~$0.5M on GB200. magic.dev/blog/pretraining
Image from Magic (@magicailabs)
9
677
Takehomes are the way to go❤️ You need to put effort in to make them fun, but there are still many things that smart humans can do which even the best LLMs can't!
Braelyn ⛓️ (@braelyn_ai) profile pictureBraelyn ⛓️
@braelyn_aiAug 20
how in the world are you supposed to conduct a technical interview when they have Fable 5
271
exabite revenue growth has been kinda crazy tbh
4
626
It's true that there's work involved in converting a raw data dump to usable training data. Yet you should still try to get a fair deal on your data! If you're considering selling your data to another data vendor, consider DM'ing us. maybe we can get you a better price :)
Aakash Sabharwal (@aakashsabharwal) profile pictureAakash Sabharwal
@aakashsabharwalAug 11
I’ve been getting this question a lot lately: "My startup is shutting down, we had 70 people, tons of data - how much can I fetch for it?" The short answer: your data right now is worth far less than you’d expect. The long answer: ⬇️
3
513
If you love and understand computers (and the web), we want you to at Exabite. We have a shiny new job board! Come find tokens that nobody else has ever seen. exabite.ai/jobs
1
1
6
15,444
We are launching as Exabite Frontier AI labs have only exploited a small % of the world's data. We've found: - 10s of trillions of corpus-quality tokens (100% incremental to common sources) - Exabytes of total OTS data - Hundreds of Petabytes of total PDFs DM to inquire
Image from Exabite AI (@ExabiteAI)
1
3
9
566
tmychow.substack.com/p/pre-t…
Jaime Sevilla (@Jsevillamol) profile pictureJaime Sevilla
@Jsevillamol19 Nov 2025
Image from Jaime Sevilla (@Jsevillamol)
5
2,257
there is a lot of alpha in actually looking at the data
Epoch AI (@EpochAIResearch) profile pictureEpoch AI
@EpochAIResearch3 Nov 2025
We looked at OSWorld, a popular evaluation of AI computer use capabilities. Our findings: tasks are simple, many don't require GUIs, and success often hinges on interpreting ambiguous instructions. The benchmark is also not stable over time. See thread for details!
Image from Epoch AI (@EpochAIResearch)
11
1,821
making a new dataset with 6 - 7 trillion tokens
1
13
998
Beware narratives!
Miles Brundage (@Miles_Brundage) profile pictureMiles Brundage
@Miles_Brundage28 Oct 2025
Seems like the whole "RL just surfaces intelligence, it doesn't increase it" series of papers is just an artifact of RL being a small fraction of compute in most LM contexts still, no? AlphaGo (etc.) shows quite clearly that there is nothing to this as a general matter
5
1,324
If you want top talent, you need to be willing to pay for it
2
666
How's that for a makeover? Thanks @joodalooped Check us out at calaveras dot ai
Image from Exabite AI (@ExabiteAI)
Image from Exabite AI (@ExabiteAI)
2
1
39
7,266
I've been interviewing so many impressive programmers who deeply understand the systems they work with, are driven, and have grit. We work on hard problems, pay good comp, and have an exciting mission. If you think you'd like it here, please DM me!
Zy (@ZyMazza) profile pictureZy
@ZyMazza25 Aug 2025
U should be required to poast the salary you’re offering when you say stuff like this
8
1,359
Us on a sunny day, making some datasets for your LLMs Only the freshest, most nourishing data for your models ❤️
Image from Exabite AI (@ExabiteAI)
3
629
join calaveras. leave your legacy in the weights of superintelligence.
Caleb Parikh (@caleb_parikh) profile pictureCaleb Parikh@caleb_parikh19 Aug 2025
I wish more job ads said things like "secure a spot in history". (from the DARPA become a program manager page)
Image from Caleb Parikh (@caleb_parikh)
1
1
8
2,086
Data is more intimate than other forms of AI engineering. You're molding the mind of the machine.
3
567
Load more