Out of the box, long-horizon agents struggle to accurately perform end to end work in the real economy (outside of coding) because those tasks are not easily verifiable, the data is hard to scale, and going from inputs to real outcomes can actually take many days. Even if you had a reliable way to verify outcomes at scale (and weren’t bothered by the multi-hour iteration loops), the sheer volume of decisions by the agent that occur in a multi-hour job makes it hard to know whether performing well will generalize to production. Over the last two years at @trybasis, we've been solving this problem by supervising the process our agents take to get to outcomes, rather than just looking at whether the outcome itself is correct. We think this is the key to building production agents at scale. It's what has allowed us to run agents in production that operate for hours, sometimes days, and reliably perform tasks like entire complex tax returns end to end. Today, alongside @braintrust, we're open sourcing a standard for defining, evaluating, and eventually rewarding agent behaviors. Thread below with all the details on how we’re scaling behaviors to close the loop for long-horizon agents.

We're building the data infrastructure that AI actually needs. Current systems were built for humans reading dashboards. But an H100 can consume 4 million images per second. The future isn't human-scale. It's machine-scale. Introducing Spiral: Data 3.0 🌀 1/8

I’m thrilled to share that @forward_deploy has raised a $45M Fund 2 for mission-critical startups! We’re backed by the Palantir Mafia: a motley crew of 150+ operators, hackers, designers, spies, machine learning scientists, classification experts, government contracting officers, nuclear engineers, and founders. Also participating: AngelList Systematic Fund, @BainCapVC, Dauntless Ventures, @garrytan, GREE LP Fund, @invariantes, @MultipleCapital, Norton Holdings, and two undisclosed large institutions 🏦🏦. We invest in startups modernizing critical American industries including defense, energy, and manufacturing — the core infrastructure that underpins our national prosperity. It’s true that “software is eating the world”, but these trillion dollar markets have had relatively little innovation. It’s time. 🇺🇸 We invest in talent mafias such as Palantir, Anduril, and SpaceX where we have strong relationships and high conviction in the people and culture. We've seen what extreme talent, velocity, and determination can achieve. Born from Palantir’s Forward Deployed Engineering culture, we back hardcore founders that move fast and run through walls to achieve their missions. We run with them and accelerate their momentum. Huge thanks to the founders and LPs who’ve given us their time, trust, and partnership. We’re fired up to build the future with you. 🔊 Turn your headphones up, this hits hard! Let’s go! 🚀

Grateful to be part of this one-of-a-kind team as I step into this new role as Partner at @AvidVc after 4.5 years of learning from the best in the business (@addielerner)! And, this team just got even better with the addition of @ZeezaCole -- I'm honored to call you my partner!

I am so excited to welcome @ZeezaCole to the @AvidVc team as a Partner and to congratulate @talivog on her promotion to Partner! Zeeza & Tali are forces of nature & unbelievably dedicated to their founders. We are lucky to have Zeeza join us & thrilled to double down on Tali :)

Hosting founders & investors in industrial tech for breakfast at @SHACK15sf on 10/10 🥐☕️🏭 Hear from Big & Small VCs: @aidanmadiganc at Eclipse @ZeezaCole at @BainCapVC @STHoward at @houseofvc @mckenefake at @ironprairievc myself at @TheCouncilCap #SFTechWeek @industrybloc

Bain Capital & Headline just released their 2024 Top 50 Emerging Vertical SaaS Companies List. The list, company breakdowns, and some highlights are included. Dive in below 👇

🎙New episode: Catch @ty_findley w/@BainCapVC's @ZeezaCole on the story behind BCV's Industrial Renaissance practice, the "new era in industrials,” and how BCV’s access to the Bain Capital network delivers differentiated value across the portfolio. podcasters.spotify.com/pod/s…

In the world of AI applications, companies are automating jobs vs. providing workflow tools for existing information workers. This has led to a set of companies (including in our portfolio) with extremely high ASPs and extraordinary growth rates. Companies are willing to pay a lot more for software that automates an entire job compared to software that makes an existing job more efficient. The question is what happens down the road. Workers are easy to hire and fire. In fact, much easier to hire and fire than ripping out workflow software. So in theory, these AI nurses, call center agents, SDRs, bookkeepers etc will also be easy to hire and fire. Companies can simply try vendor A today and switch to vendor B tomorrow if they provide an automated agent that is cheaper, better, or faster, potentially leading to massive price compression over time. [Look what happened to the BPO space] Models can always be improved on and the quality of a model is a tough thing to rely on for defensibility. In the early days of website personalization, many startups got formed to provide recommendation engines ("if you like product A you will also like product B"). These engines initially took off and provided massive lift to the ecommerce sites...but over time, the websites would bake-off these incumbent vendors against new entrants. If the new entrant was better, then the ecommerce site would switch. As a result, despite the value delivered, this space never became huge because there was churn and price compression. Will this happen with this current crop of AI software. Unclear. Personally, I don't think defensibility will be based on training data or quality of the model/engine. Ultimately, the best of these newer companies will win because they are deeply integrated into their customers' workflow and because their solution is being touched or interfaced with lots of information workers on daily basis.

Friday Props: The defense, manufacturing, and deep tech ecosystems unequivocally owe a debt of gratitude to my good friends @MikeSlagh @austinbishop @aphysicist and the many others who turned @newindustrials into reality. At a moment’s notice, they brought together hundreds of tenacious and energetic people from across the globe to confront an existential risk. We don’t have decades to innovate. As @2112Power emphasized, America has three years to fix what’s long been decimated and ignored. Luckily, this is what we do; this is where we thrive; and it’s abundantly clear many exceptional builders, investors, bureaucrats, and politicians aren’t just doubling down on deep tech, they’re going all in. What an inspiring two days! Shoutout to those actively supporting the cause. It’s always a privilege to catch up with @ConnorLoveCA @MichelleVolz @michaelxbloch @AlexaLiautaud @benkohlmann @Dan_DePaoli @ZeezaCole @vc @TaylorCSargent @marvinliao and many more. Same is true of the builders at the core of this renaissance: @FilArons @williewockets @JoshuaSteinman @CameronLMcCord @ssankar @naweed @wyatt_h_smith and many others.

See 2024’s Top 50 Vertical SaaS Startups ✅ IN: AI transforming industries like … Construction tech | Legal services | Logistics 🚫OUT: Old-school systems | Paper-heavy workflows | Fragmented software baincapitalventures.com/insi…

Heading back to where American manufacturing and industrialization began: Detroit. Excited to see so many good friends @vc @MichelleVolz @TaylorCSargent @williewockets @CameronLMcCord @michaelxbloch @ConnorLoveCA @mpalank101 @naweed @austinbishop @ZeezaCole @benkohlmann. Massive shoutout to @MikeSlagh @aphysicist and team for hosting Reindustrialize! Looking forward to many more years of meaningful gatherings!

@BainCapVC & @HeadlineVC second annual Vertical SaaS 50 list: @SquintAI is my favorite so far...

Basetwo is named one of @BainCapVC’s 2024 Top 50 Vertical SaaS companies - recognized for helping engineers boost efficiency & reduce cycle time with AI! Big thanks to @Hub71AD, @aeroxelerated, @GlasswingVC, @PlugandPlayTC, & @ArgonVc for your support! hubs.ly/Q02BpnHT0

Today, @BainCapVC and @HeadlineVC recognize 50 emerging vertical SaaS companies with our second annual Vertical SaaS 50! Read more here: baincapitalventures.com/insi…

"Vertically integrating is challenging, but by building an end-to-end solution, companies can offer a comprehensive, and by extension a stickier, solution to their customers." - @ZeezaCole @ajay_bcv 🙌🙌🙌

Over the last 20 years @BainCapVC has had the opportunity to partner with founders transforming the physical world. Today, @ajay_bcv and I are excited to launch our Industrial Renaissance practice to continue to find, fund and foster the best talent building for industrials.

Thank you to all who came out to the Riot Ventures office for an evening of tacos, margaritas, and early-stage builders! 🛠🌮 If you are a HardTech founder or curious about the LA HardTech startup scene and seeking community, join us quarterly to gain exposure to early-stage founders who are a step ahead and may be able to provide you insight on the process. Hosted by: @meghan_moreland (Riot Ventures), along with @ekhenriksson (@rootvc), @ZeezaCole (@BainCapVC), and @alejandravl13 (@BeePartners)

Last night’s Vertical SaaS investor dinner - our biggest one yet! Hot topics of the evening were: - tech enabled services - healthcare back office - second wave of legal AI startups - saas for insurance - CAD software landscape …and much more! Have had fun hosting this series. Dm me or @urvashibarooah to join the next one :)