Tracy Liu

Investing @M12VC / @uchicago

[Tracy Liu @TracyyLiu]

We are hiring a senior portfolio development manager! Lmk if you are interested!

https://t.co/f0rQqx4A6i

[Tracy Liu @TracyyLiu]

@styskin @ArtificialAnlys @KeenableAI Congrats!!

[Tracy Liu @TracyyLiu]

@edi_sf @ElevenLabs @scalevp Congrats!!

[Tracy Liu @TracyyLiu]

@buccocapital Long meta

[Philippe Mallette @PhilippeMallett]

@SAP buying @prior_labs might be a bigger financial services story than it looks.

Prior Labs isn’t a fintech company. But its work on tabular foundation models points to where a lot of AI value in financial services may actually be created: deeper in the infrastructure layer.

Banks, lenders, and payments companies already rely on ML for underwriting, fraud, risk, and transaction monitoring. The problem is that many of these models are still built one at a time, with long development cycles and costly overhead.

TFMs could change that.

They’re not LLMs for spreadsheets. They’re models purpose-built for structured data, with the potential to make predictive model development dramatically faster and more scalable.

What LLMs have done for software engineers, TFMs may do for ML engineers.

That matters for expanding credit access, launching fraud systems for new products, and giving stretched ML teams much more leverage.

But in financial services, performance won’t be enough. Explainability, auditability, and regulatory trust will matter in getting into real enterprise workflows.

My colleague John Onwualu and I wrote more about this here: https://t.co/n6E7Er7VBY

If you’re building in this space, especially if your product isn’t “fintech” on the surface but could fundamentally reshape how financial institutions operate, we’d love to connect.

[Tracy Liu @TracyyLiu]

Microsoft Ignite by day. The Hunger Games by night. 🎬 Join us for a special pre-screening to close out the week!

RSVP: https://t.co/09xwDHQj2X

@M12vc @Microsoft

[Tracy Liu @TracyyLiu]

@gauri__gupta Congrats!!

[Tracy Liu @TracyyLiu]

@NousResearch Yes

[Nous Research @NousResearch]

Happy NousCon day to all who observe

[Tracy Liu @TracyyLiu]

@sonofalli @droid @FactoryAI @dee_bosa @theashleyjepson @annbordetsky @francesca_lab @cjc @NancyZWang @maddiehfaulkner Ty for hosting!

[Tracy Liu @TracyyLiu]

@jamesfly @sequoia @getserval Congrats!!

[Tracy Liu @TracyyLiu]

@denise_teng25 @plato_rl @GradientVC @darian314 @RobFarlow @pranav__putta 🚀🚀🚀

[Tracy Liu @TracyyLiu]

@annbordetsky @FactoryAI @dee_bosa @cjc @NancyZWang @theashleyjepson @francesca_lab @sonofalli @eliza_bara @maddiehfaulkner Incredible!!!

[Tracy Liu @TracyyLiu]

@PhilippeMallett Been getting too many spams from agents

[Tracy Liu @TracyyLiu]

@joeschmidtiv @hxtweets @amrit_hx @astrange @santiago__rdz Congrats!!

[Tracy Liu @TracyyLiu]

Halloween M12 Edition

@P_Lenke @protokultur @alessandrolev @M12vc

[Tracy Liu @TracyyLiu]

@gauri__gupta 100%

[Tracy Liu @TracyyLiu]

Elizabeth Holmes raised ~$1B for Theranos....the uncomfortable part is how much of VC is just that - believing the story before the evidence exists :)

[Tracy Liu @TracyyLiu]

@joshua_sirota 🚀🚀🚀

[Philippe Mallette @PhilippeMallett]

Anthropic has joined the personal agent space, but where are the long-assumed winners of the personal context game, Apple and Google?

I’m bullish on the non-platform agents because the biases that every one of these agents will inevitably have as they seek monetization will be influenced by the platform around them. Meta wants to own more of your mindshare and behaviors to get more from their platforms, so eventually it just becomes your way to access the walled Meta garden. Google and Apple already know and monetize users well; connectivity to other platforms will be de minimis to them so no need.

Dots, Grokbot, Duet all know the subscription business well, and their parents will bias them to have lower usage and likely a worse experience as they evolve so you stay on the main subs. A free version of these agents cannibalizes a lot of that market (which also serves as a fantastic PLG motion towards an enterprise subscription), and they’ve acknowledged as much by broadly skipping the free version.

That leaves Instinct, @wajo_ai, @hark_labs, and the hundred other independent personal agents. You won’t pay a subscription fee, they can’t upsell you to an enterprise license, and they don’t have the ecosystem to indirectly monetize off you.

So what’s their bias? Financial services. To survive, they likely need to win your payments, banking habits, investments, insurance, and beyond. So they’ll be incentivized to do what everything else in the world already does: make it as easy for you as possible to use your money. And if you already do that, it’s the least disruptive bias.

I’m bullish on the independent agents. Everyone’s biased, but their biases are the most palatable.