Partner @BatteryVentures. Views are my own, no investment advice intended.
NYC
- Tweets 2,471
- Following 1,400
- Followers 1,293
- Likes 4,539
Sparks of Artificial General Intelligence: Early experiments with GPT-4 (March 2023)
opus 5.5 can rap now. here's the first rap single and music video: "No Samples" 🔊
everything you see and hear is powered by custom javascript code written by opus. confused? don't worry, claude raps about how it all works
6
263
Counterintuitively, the best way to break caffeine overconsumption is to have twins since you forget to drink coffee.
I’ve gone from 4-5+ cups a day to 1-2 during paternity leave.
2
6
455
Brandon Gleklen retweeted
Bill Clinton in Shanksville, PA, on the 10th anniversary of September 11th.
2
84
498
200,686
What is tough is that the rules of the game can change quickly, and sometimes that change can feel arbitrary (or hype-driven). And rules can change faster than you can adjust operationally.
A primary job of the board is grokking the rules and projecting how they’re to evolve.
One of the biggest mistakes I made as a founder, and I see some other CEOs making it right now, is missing when the outside world changed the rules of the game.
Every company starts by finding their point of view on "the game" (the market) and the right way to play it.
Then you go heads down and focus on the scoreboard. In SaaS, it was ARR or Rule of 40. In consumer, it was DAU/MAU.
Quarter-after-quarter, year-after-year, you grind against that scoreboard. This is the way.
But what happens when the rules outside your company radically change?
Take "Rule of 40." That was a game played in the later innings of SaaS' hegemony, when the market wanted profitability mixed into growth. As @jasonlk wrote recently (link in comments), Rule of 40 is now a dated concept. Growth is what matters again. And the expectations on growth are enormous.
For native AI companies, this is Captain Obvious. Yet I meet so many CEOs, from SaaS and other sectors, still grading themselves on the old scoreboard and trying to layer AI on top.
The hard part: you have to anticipate the scoreboard for wherever you're trying to end up.
Amazon anticipated that the short-term earnings scoreboard in the early 2000s was something they could power through. Palantir fought the perception that services were "bad" on the software scoreboard and revolutionized the category.
If your goal is to sell the company in 2 years, what scoreboard will the acquirer be using?
Selling to PE? There was a brief window (2018 to 2023) when growth was king. Now stickiness, especially with debt in the deal, is the top filter.
Selling to a mega-tech? Your EBITDA is a rounding error to them. They're buying a wedge into a market or a team.
This is why startups are hard. You have to play the game as stated. And you have to anticipate the new game coming down the road.
And most of all, you have to stay in the game.
1
1
7
815
Tools + context continue to be the biggest bottleneck for making AI productive in a general business context (we crossed the threshold on raw intelligence a while ago).
@Altimor and team are obsessive around how to bridge this gap. Excited for this to be out in the wild now.
Today, we kill the AI agent and introduce the AI employee: Lindy Teammate.
It’s just like working with a real employee.
Everyone on your team can simply hit it up on Slack and get 10x more done.
Lindy also keeps learning, and becomes the self-updating brain of the whole business.
Live now: lindy.ai
5
4
419
The ephemerality of product/market fit is one of the biggest changes in B2B investing since I started at @BatteryVentures in 2015.
Companies can have great spikes in new bookings with happy and excited customers. Retention on the logos you land can be fine if you're addressing a sticky workflow. The more common issue is your ability to book new customers at the same rate.
People frame the risk as competition, or incumbents "embedding AI." The bigger risk is that the next buyer never goes looking for a tool at all because the ChatGPT and Claude subscriptions they already pay for serve the use case well enough.
There's a threshold for pain needed to go through procurement for a new tool, and your existing subscriptions are eating away at that pain over time.
Of course there are plenty of AI native companies that are showing consistent linear (or exponential!) new bookings growth. Those are exciting companies to be a part of and the venture community is rallying behind them. But generally, past growth rates have never been less correlated with future growth rates in my experience.
Great to chat with @lauramandaro - interesting time to be an investor!
Today from me in Dealmaker: mutual fund markdowns on software startups like DataRobot and Airtable illustrate the pressure on founders and their investors
theinformation.com/newslette…
1
4
517
If Anthropic wants the public to use AI more ambitiously they should just create an addictive web game that is only playable during inference.
The only way to get a high score in AI Flappy Bird is to ask something that makes the AI think long enough.
1
3
129
A thought experiment I’ve been using recently:
If all Anthropic employees worked at your customer/prospect, and focused on your use case, would they still pay for your product?
(I think many startups pass this bar.)
1
106
Competence as Tragedy: crowprose.com/blog/competenc…
129
Excited for @BatteryVentures to lead the $22M round for @Triomicsinc and to join the board.
Oncology is one of the hardest areas to deploy AI due to the complexity of patient records (can be 300+ pages and 80%+ unstructured).
1
1
3
348
Triomics is using AI to reconstruct the chronology of care and structure the patient record.
We're live at places like MSK, MD Anderson, Yale, Mount Sinai and Texas Oncology, automating workflows ranging from clinical trial matching to registry reporting and visit preparation.
1
155
Excited to be working with Sarim, Raj and team alongside our friends at @lightspeedvp and @NexusVP to change cancer care for the better!
85
High impact role 👇
Spent a significant amount of time on our board call today talking about what an impact this person will make. Highly recommend working with @Altimor
We're hiring our first Growth Product Manager at Lindy.
Lindy Assistant is growing fast and users positively rave about it — looking for someone to help us beef up the growth engine.
DMs open!
1
269
“Application software has never had great moats”
That’s true. But there’s always been a need to procure a solution.
That changing - people self-serving their use cases using Skills, Cowork etc. - is the issue that will impact growth rates more than “undifferentiated tech.”
73
A useful definition of “AI native”:
"Inference is our #1 cost by a lot (more than payroll)"
Maybe not true for every 'AI native company' but any company with this economic profile deserves the moniker.
We've tested new OSS models the moment they're released for a while at Lindy. Inference is our #1 cost by a lot (more than payroll) — cutting it by 2-5x would be transformative.
Last year, OSS models were "not even close."
3 mos ago, "almost there." Came close to making Kimi K2.5 our default.
I think we are right now crossing the line to "at the frontier, for most use cases." GLM-5.1 in particular is incredible and will likely be our default soon.
Surprised by this development — OSS caught up.
2
304