Mark Attar

mts @FactoryAI work on important problems prev @tesla

[Mark Attar @mark_attar]

@theSciTechGuy This thing has some sauce fs

[Mark Attar @mark_attar]

@gabepereyra Even just building that shared platform such that it works for all of these F500 companies and the law firms and gets the work off email sounds like a more painful challenge than most people realize

Then the training adds another layer for sure

[Mark Attar @mark_attar]

I wonder if Axiom or Harmonic's business model in the future would be like for X dollars we'll do proprietary frontier math research in this problem space for this hedge fund or cryptography business or something that unlocks a powerful algorithm that didn't exist before

[Mark Attar @mark_attar]

@willdepue but crypto/stablecoin is also honestly the most efficient way for ai to transact in the future as the throughput goes even more through the roof

[Mark Attar @mark_attar]

@sanjayamirthraj @fliptexts @a16z @speedrun @notolegv @ajiang_xyz Congrats @notolegv! An absolute beast very excited to see where y'all go with this

I also gotta try this ive been looking for good ai financial apps for a while

[Mark Attar @mark_attar]

@shcallaway Hugely agree with this, tech should always be meritocratic and open to people from all backgrounds

Thats where the super talent usually is anyways

Part of moneyball is finding talent from every source with no dogma or fixed priors though

[Mark Attar @mark_attar]

@rhythmrg 100% agree, our thesis as well

[Mark Attar @mark_attar]

@appliedcompute @nvidia Awesome stuff

[Mark Attar @mark_attar]

@appliedcompute @nvidia I see what you did there

[Mark Attar @mark_attar]

@shevchenkoaalex @Neel490 @avikrishna To be clear not being a neolab is a good thing, neolab means you are pre revenue and ramp is very very not pre revenue lol

[Mark Attar @mark_attar]

@avikrishna No they’re good at being on the cutting edge and doing research it’s just that neolab means pre revenue and ramp is very not pre revenue lmao which to be clear is a very good thing

[Mark Attar @mark_attar]

@avikrishna They’re not a neolab they have so much revenue lol more like just a fantastic applied ai team that knows how to innovate and experiment

[Mark Attar @mark_attar]

@_sholtodouglas @livgorton Yeah it’s amazing I do still feel my human bottleneck weighing, and even still feel like I have many more ideas than I have execution ability despite many concurrent threads which I didn’t expect yet at this stage, this will certainly change

[Mark Attar @mark_attar]

Ramp might just pip basis and rillet and campfire to this accounting thing honestly didn’t realize that was possible

[Mark Attar @mark_attar]

@harvey @FireworksAI_HQ Wow that is some crazy cost gap

[Mark Attar @mark_attar]

@gabepereyra @FactoryAI @matanSF Wow that’s awesome

[Mark Attar @mark_attar]

@sqs Rebasing my worktrees on dev or performing simple refactors do not require Opus and cheaper open source models do a fantastic job saving both time and money in these cases especially when model router takes care of the mental overhead of switching

[Aaron Levie @levie]

As token budgets take on a larger part of operating expenses over time, model routing is the inevitable conclusion. This is also one of the biggest areas of differentiation for the applied AI layer over time.

By understanding the different work patterns in your domain, and having strong evals for that domain, you’ll be able to cost/performance optimize effectively.

We’re still likely at the point where most use-cases will need frontier performance for the foreseeable future; but soon you will be able to peel off individual use-cases and send them to lower cost models once the quality is sufficient for the task.

Enterprises individually trying to figure this out themselves at scale will likely not be possible, so the products that can intelligently route these workflows to the right tier of model will be in a strong position to aggregate more demand.

[Mark Attar @mark_attar]

The frontier labs will make the best models, but they won't be able to make a model for every package of price, speed, and ability

Thats where being an aggregator becomes an advantage

[Mark Attar @mark_attar]

@FactoryAI Shoutout to our boy @_AbhaySinghal for cooking this up