Will Meinhardt

Head of Sales at Mach 1

[Mach 1 @mach1ai]

Craig McGowan manages the Mach 1 agents at Trace.

Here’s why he loves Mach 1. A few highlights:

> “The flexibility is kind of insane.”

> “If you need something that is not AI-focused but you’d like to use AI to build it, you can do that too.”

> “It’s always up. I’m trying to think if I’ve ever worried about it being down, and I haven’t.”

> “I love how easy it is to talk to everyone at Mach 1... Everyone I work with there is fast at responding and nice.”

Full interview with Craig and @rmeinhardt coming soon.

[Mach 1 @mach1ai]

We tested five models on the same customer support workflow in Tower, then asked each to update it when the refund policy changed.

All five passed all 10 test cases. The cost to build and update ranged from $0.48 to $15.61.

Full results below. https://t.co/Ekd2NOMCoY

[Will Meinhardt @w_meinhardt]

@shubgaur Let's go!!

[Will Meinhardt @w_meinhardt]

The story behind the parable: https://t.co/MoMUrVxZZd

[Will Meinhardt @w_meinhardt]

If you had 4 months to make the best pottery you could, should you focus on quantity or quality?

In the Art & Fear parable, the students graded on quantity(!) made better pots than those graded on quality, meaning they learned by making more.

It’s also important to learn the basics from people who know what they’re doing. Without any guidance, you can spend unnecessary effort figuring out things someone could have shown you (like which clay to use in the first place).

This is how we approach AI deployments. We thread the needle between giving our customers some direction (infrastructure and best practices) and also giving them the freedom to explore what works for their business without losing what makes it unique.

[Mach 1 @mach1ai]

You can manage your Mach 1 agents through Grok @bot.

In this video, we update the pricing used across our workflows by uploading a simple .md file with the latest information.

[Will Meinhardt @w_meinhardt]

I just walked to our office and listened to a custom 31 minute .mp3 about every sales call I have today.

This is amazing and it costs less than $1/day in LLM tokens. If it cost $200/day to hire a human to do this, I never would have paid it.

Some basic economics: this doesn’t mean I’m GETTING $200 of value by automating the work with AI. I'm getting somewhere between the token cost (less than $1/day) and what I refused to pay before ($200/day).

One of our favorite phrases at Mach 1 is, “AI allows you to pursue opportunities where the economics of doing so didn’t make sense pre-AI.”

Some work would have added value, but not enough to justify paying someone to do it, so it never got done.

When choosing what to pursue, we recommend starting with the work you ALMOST would’ve hired someone to do, but where it didn’t quite pencil out. Then work your way toward the things that were harder to justify. Some still won’t be worth the cost, even with AI.

Customer success is a good example. We have customers who now use AI to handle customer success interactions for smaller accounts paying less than $20k/year. Those accounts get more attention, churn less, and buy more.

ALSO, sometimes AI finds an upsell or a large deal that changes the math. The company thought the work wasn’t worth paying for, yet once it became cheap enough to try, they found out it was.

Either way, add enough of these wins together and they can compound into a significant competitive advantage.

[Mach 1 @mach1ai]

Our latest benchmark results are in.

We tested four models from OpenAI, Anthropic, DeepSeek, and Google across 38 orchestration scenarios ahead of a Mach 1 release coming in the next few weeks.

Full results below. https://t.co/x4ymKZA1yU

[Rob Meinhardt @rmeinhardt]

Why do customers love Mach1? Listen and learn.

Thank you Praktika!

#mach1 love

[Rob Meinhardt @rmeinhardt]

Go hands-on or you won't see it...

#Mach1ai

[Will Meinhardt @w_meinhardt]

We’re hosting Steaks & Systems in Chicago on Thursday, September 24.

Samuel Ouzounian, our lead forward deployed engineer, and I are bringing together a small group of operators for dinner with David Friedman, CEO of Panther Capital and a Mach 1 customer.

A nice dinner with good people and fellow operators, with some practical conversation about getting value from AI in business operations.

If you’re an operator and would like to join, shoot me an email at [email protected].

[signüll @signulll]

hinge should’ve launched a personal agent that automates your dating life.

btw this is basically what i would’ve done with bumble if i were ceo.

there was a real opportunity to turn the entire platform from a marketplace of profiles into a tasteful mechanic that personally introduces you to people, helps you connect, coordinates the date, & stays involved afterward.

the agent could ask how the first date went, understand what you liked or didn’t, mediate whether there should be a second date, & gradually learn who you actually fall for.

surprisingly no one has done this tastefully yet.

[dar @radbackwards]

Everyone everywhere feels like they’re missing out on something right now.

The New Yorker feels they are missing the tech renaissance

The SF kid thinks they’re missing the equity gains of the big labs

The open ai employee thinks they’re missing out on life as it passes them by

Avoid the FOMO and bet the house on whatever makes you happy + puts love in your heart and do it extremely well.

Everything else will sort itself out.

[Mach 1 @mach1ai]

Tower lets operators manage agents and workflows from one place, with clear visibility into what's running and what needs attention.

[Rohan Paul @rohanpaul_ai]

"The episode of history where humanity was the apex intelligence is about to end."

AI is crossing an important boundary: from learning mathematics invented by humans to inventing mathematics humans can learn from.

Mo Gawdat, ex-Google executive explains how recursive self-development of AI is absolutely disrupting mathematics.

Humans originally built AI using human intelligence, but increasingly, AI can can improve itself (i.e. recursive self-development)

Once machines become capable of making increasingly better versions of their own methods, technological progress may accelerate far beyond the normal pace of human research.

----

From "Mo Gawdat" YouTube channel, (full video link in comment)

[Will Meinhardt @w_meinhardt]

@robcgiani Badass video

[The Figen @TheFigen_]

The real danger is not that computers will begin to think like men, but that men will begin to think like computers.

- Sydney J. Harris

[Finding Compounders @F_Compounders]

A young Warren Buffett

[Will Meinhardt @w_meinhardt]

@salesxsaas @mach1ai to leverage both non-deterministic and deterministic workflow steps. Often you need a combination but you shouldn’t have to be the one tediously building it out on a flowchart.

[Will Meinhardt @w_meinhardt]

@andrewdsouza @boardyai @mach1ai @andrewdsouza we appreciate it!

[Will Meinhardt @w_meinhardt]

We’ve won multiple deals against Anthropic in the last month.

Anthropic is a trillion-dollar company on the verge of an IPO, with undeniably great products. @mach1ai is a much smaller company.

So how do we win?

1/ We are built specifically for operations. In many of these deals, the engineering team is already using Claude Code, Cursor, or Codex. That doesn’t change. Those companies realize they need something different for operations. They don’t want their engineers spending valuable time building and maintaining internal tools. Our secret sauce comes from our founders’ deep experience running businesses. They know what it takes to build systems that work across teams, handle messy edge cases, and hold up in production.

2/ We give companies the infrastructure they need to avoid reinventing the wheel and the flexibility to preserve what makes their business unique. Customers don’t want to be pigeonholed into one model or a narrow point solution. It feels like beating a dead horse, but the ability to switch between models is key. When Anthropic went down in July, we were able to move affected workflows over to other model providers and keep our customers running.

3/ We stay extremely close to our customers. That’s not to say frontier model companies don’t care about their customers. But we work with roughly 25 enterprises, so every single one matters enormously to us. We benefit from using AI internally without carrying the huge overhead of training frontier models ourselves. That allows our best people to work directly with customers.

If you're thinking through how to get value from AI in operations, send me an email at [email protected] and we'll set up a time to talk.