Hank Couture

@ElisaSchreiber @felicis Great job. Everyone at Felicis has built an incredible firm the last 20 years. Look forward to seeing the mission continue

Founding a company is an act of courage. Today, we launch a fresh new look for @Felicis built on the core belief that has guided us from the beginning: Fortune favors the brave. 🔗 felicis.com

@omooretweets Very much agree

A big story that most people are missing in the AI race for the consumer (ChatGPT vs Claude) is ads. Right now, most consumer AI revenue is coming from power users who are willing to pay high cost subscriptions. This currently skews positive for products like Claude - but this will not be the end state. Google makes ~$460/ user/year in the U.S., mostly on ads. Meta makes around ~$250. I would argue ChatGPT’s ad-based ARPUs will be even higher as they will ultimately have deeper / more frequent user engagement. Even at the $460 level - monetizing everyone in the U.S. via ads is $152 billion in annual revenue. By contrast, if you’re able to monetize even 5% of the population on a $200/month subscription (which is a stretch!), that’s only $40 billion 🤔 I suspect this will be even more drastic outside the U.S. where users are even less willing or able to pay directly for subscriptions. And, the earliest data from a very small rollout shows ChatGPT ads are already outperforming Meta in effectiveness - this just gets better over time. TL;DR - I would not count ChatGPT out on consumer AI revenue. Once ads start working, that can quickly become a massive machine.

congrats to Theta Neurotech on the $1.75M pre-seed to literally save lives with AI Truman + Chris were LeapYear 2024 and worked out of the office with me for a few months. I can think of no team more relentless than the two of them

@credistick @credistick would be interesting if there is data that does a like-for-like comp of startups going public vs staying private as they mature. Like yrs 5 thru 15 50 companies in each bucket. Similar amounts of funding. Compare buckets Perhaps not large enough sample size yet

NVIDIA went public at a market cap of about $625 million, and became a $4.42 trillion monster in public markets. This example does not support a thesis of much larger exits, rather it’s an argument for going public sooner. Companies have been going public later (and larger) because VCs wanted to extract rents on private market growth when interest rates were low. Not because it was better for founders. Exits will generally be larger now than in previous eras, but much of the current activity is downstream of ZIRP conditions and will not persist for the next generation. Public markets provide more meritocratic access to huge pools of capital, which will look increasingly attractive to the best companies. And that’s a great thing, for the best companies and their investors.

State of vibe coding right now

There are so many companies tied up in useless meetings and 1:1s all day While competitors are shipping quickly with the latest AI updates out of Claude, Codex, etc. Lots of alpha at the frontier

@literallyhimmmm Claude api in cursor

@HankCouture What have yall been doing the last 12 months instead of this

overdue but finally switched from Cursor to Claude Code yesterday. 24 hrs in. Fully Claude pilled. Feels like I up-leveled from engineer to eng manager…an orchestrator of agents

@dhasandev I agree with these. Good stuff. Will check out Squad

soon you'll find that orchestrating teams of agents requires two things: - long term context engineering (your team will get confused, and you will as well) - dynamic orchestration (youll work with 4 different agents on different things, then you'll want to create one parent agent to self-drive them for you) we're building this at trysquad.ai and would love to hear how you're building with your agents :)

@alexfbowden Won’t tokens be more expensive? Big reason I switched to Claude code was cost. seemed higher ROI to go direct

@HankCouture Use terminal in cursor, then you get Claude code, codex, Gemini, and cursor all in one

@joshuaday Inching my way there but I can see it. Basically run a whole Eng team at once

@HankCouture wait until you start running 3-4 sessions at once in different worktrees. that’s when the orchestrator feeling gets real. one agent building, one reviewing, one writing tests

@mylesmarino Oooh interesting. Will check out

@HankCouture You should try conductor. Add steroids

@Ncky_Mrtnez 😂

@HankCouture Welcome to the club. The "orchestrator of agents" feeling lasts until Claude decides to rewrite your entire project without asking

@budwindruid It’s really unreal

@HankCouture so worth it

The cost of building is going down. A good builder is more valuable than ever

@CassiusDeHaight @toddsaunders Back in the good old days!

@HankCouture @toddsaunders Damn shame this wasn't done over a picnic table and some coffee

spent this AM riffing live w/ @toddsaunders where b2b software goes next 5-10 yrs - - in time, every business will have custom software for its day-to-day workflows - value capture is either in owning the inputs or the outputs to that scenario - inputs = data or infrastructure that feeds into that custom software. e.g. the HVAC business might have custom software to manage its day-to-day but is not going to build custom payments infrastructure. They could try but the gain does not offset the risk of messing up payments or missing fraud detection - outputs = while anyone can claude code, they might not know what to prompt. Just like anyone can create a TikTok but some people are better than others at it. So if i'm the HVAC business, i might still want to hire someone to build the custom software for my needs. I can either hire that person directly or hire a "claude code agency"? Historically, this has always been possible but its been expensive. It's going to get much much cheaper. The counter argument is the LLM conversation could be so good at user discovery with the HVAC operator that it can basically become an agentic PM and perfectly craft the solution itself without a middleman. I'd bet that does happen....but also humans still make many decisions on trust and expertise. For example, many businesses pay McKinsey millions for its expertise. Is it better? idk but if you start a contract with it, you likely trust it is better and that trust helps you get comfortable with the final outcome. If an AI tells you this is the right answer, you might trust it. But you might feel better with a Prompt Engineer expert telling you this is the right answer. In reality, we will probably see both options for a business owner. The cheap option, which is trust the AI product manager and the expensive option, trust the "McKinsey Prompt Engineer" to get it right for you. I use McKinsey purely as an analogy for a smart and trusted person / brand. It could be McKinsey, but imagine there will be an entire market of services and names that do this --> humans delivering the output of custom software that is trusted and cheap Now - the other counter argument would be "many people are lazy and still would rather use and trust an off-the-shelf vertical software for their business". I think that could be right...but with the way things have evolved the last 5 years...I'm not sure I'd want to put too much weight in that argument. The risk of being wrong outweighs the upside of being right. I think you'd want to build for both scenarios...e.g. provide an off-the-shelf vertical software solution anyone can use but empower them to also plug their custom software into it easily. If you own the inputs to the custom software, its just a UX swap and shouldn't really impact your business at all. It's providing the customer-first solution of "you pick what you prefer. our interface or your interface" note - i wrote this in 15 min just downloading thoughts after call with @toddsaunders. Likely many sub-factors and derivatives we didn't think through