Talk to me nice. Building https://xitter.cf/t.co/NQ3OQREh7v
Los Angeles, CA
Joined December 2011
- Tweets1.9K
- Following754
- Followers786
- Likes14.9K
bmax retweeted
Bill Gurley, Jack Altman back startup Pursuit, which helps companies sell to government techcrunch.com/2026/04/29/bi…
bmax retweeted
My goal when I became a full time investor was to be the type of partner I would have most wanted during my time building a company: someone who had done the work I was doing before, who had the network to help me reach my goals faster, and who had my back no matter what.
I’m really happy to share that we’ve raised Alt Cap II, a $275M early stage fund.
We backed ~20 companies at Seed and Series A in Alt Cap I. The new fund will be the same structure with a bit larger checks. But mostly we just want to be in business with people who inspire us.
There’s no way to say it without sounding cheesy but it’s true; I feel unbelievably lucky to do this job in the way I get to do it.
A special thanks to the founders that let us back them in Alt Cap I, you all are what it’s all about for me!
Nice
Episode #92: How AI Changes Governments & Business Models
@MikeVichich is the Co-founder and CEO of Pursuit, which helps companies drive more revenue from the public sector.
We talk about how AI is changing Helmer’s 7 Powers, how it’s impacting the government, if DOGE is actually working, building a startup in the Midwest (specifically Ann Arbor, MI), and how to disagree with your team.
We also get into Mike’s prior company Wisely, and how they went from $11 in the bank account and unable to run payroll for six months, to over $10 million ARR and a $187 million exit to public company Olo a few years later.
Thanks to @jaltma and @Blakeir for their help brainstorming good topics for Mike!
Full episode here on X, or grab a link in the replies.
Timestamps:
3:21 How AI changes Helmer’s 7 Powers
17:06 What becomes important in AI-first economy
21:02 How AI interfaces with the government
24:02 “The rules intended to save taxpayer money ironically cause taxpayer money to be wasted”
29:34 How change orders impact public sector costs
33:20 Why DOGE has not impacted US government spending yet
38:15 Three pieces of wisdom from 2nd-time founders
41:44 Starting Pursuit to make selling to the public sector as easy as the private sector
45:35 Why cities grow expenses 5x faster than tax revenue
51:42 Pros + Cons of building startups in Ann Arbor, MI
57:43 Hiring talent density in the Midwest
59:30 Starting his first company to fix consumer credit cards
1:08:50 Pivoting Wisely to restaurant loyalty
1:12:49 $11 in the bank, missing payroll for six months
1:15:21 Embarrassing demo at an Ann Arbor tech meetup
1:18:18 Why CEOs don’t always have to be right
1:20:54 How to disagree
1:25:48 Hiring at Pursuit
1:28:30 “A bad day with customers is better than the best day in the office”
1:31:33 Crashing their first customer’s PoS on Labor Day Weekend
1:35:55 Using “The Cadence” to hit $10M ARR
1:41:55 Selling Wisely to Olo for $187M
A pleasure working w/ y'all
Case Study: Learn how @PursuitGov transformed their B2G offerings using LlamaParse:
➡️ Parsed 4 million pages in a single weekend
➡️ Increased accuracy by 25-30% for complex document formats
➡️ Enabled clients to uncover hidden opportunities in public sector data
See how LlamaParse helped Pursuit create a searchable database of public sector documents, empowering B2G sellers to identify new initiatives and funding streams:
llamaindex.ai/blog/pursuit-t…
The new structured outputs OpenAI released is pretty awesome, and, should help a ton of people.
Preliminary testing makes it not useable yet though:
- gpt 3.5 turbo .7s -> 13s w/ strict
- gpt 4o mini 1s -> 9s w/ strict
This is how I've felt evaluating every AI tool or platform out there. Consistently. It's exhausting, and being a good citizen by taking the time to make a PR or call out the problems you find doesn't solve the immediate issue or make you want to pay for the product!
Ok because I'm in the market for LLM ops/gateway for @smol_ai, i just spent part of my day trying 3 of the (word of mouth approved) options. here's how it went:
- code copied from docs dont work (sdk api changed lol. easiliy fixable but sloppy)
- code copied from docs results in uninformative error codes requiring me to ping support
- JS SDK adopts Python naming conventions (this-is-annoying vs thisIsJSNative)
- $49/mo for storing and concatenating prompt templates for me? (i know $MAJORPLAYER1 charges... more too haha)
- $MAJORPLAYER2 does not let you set budget/rate limit for keys?
- nobody has implemented short lived keys yet???
- everyone does fallbacks (dubious?) and status code based retries but nobody lets you set timeout?
- i need higher granularity than just "user" or "project", what about "project-feature-user"? can i facet the dashboard? no? fml
- nobody supports a dspy-like loop
@AlexReibman says there are like 80 options on the market, i dont have time to eval all of these. does anyone want to work with me to do a Wirecutter like review of implementing the same app using the top 10 (you pick with me) and publish on @latentspacepod?
bmax retweeted
High agency has many great upsides, although a significant downside is that it is also highly correlated with high anxiety.
bmax retweeted
"I grew up implicitly thinking that intelligence was this, like really special human thing and kind of somewhat magical. And I now think that it's sort of a fundamental property of matter..." @sama