I wrote a long essay following a talk I gave at AI DevCon in Brooklyn last month.

It starts out with an anecdote about hunting ChatGPT in December 2022 and goes on to explore what I think will be necessary to thrive as code becomes a commodity:

In December 2022, I hunted ChatGPT on Product Hunt.

It ranked #1 product of the day, then the week, and went on to be named Product of the Year.

Having co-founded a YC-backed conversational AI startup in 2018 (long before LLMs) I recognized in ChatGPT the missing ingredient that would have made that venture viable.

The future we d anticipated had arrived. I could revisit my old problem, or I could expand my area of potency by raising and deploying my own venture capital fund.

I chose the latter.

Three years later, on December 9th, I watched a 24-hour window on Product Hunt cross 500 launches roughly double what I observed throughout the preceding 825 days. Only 13 were featured; most were unremarkable.

The LLM has fundamentally shifted the economics of software development.

As someone with a dual vantage point being the #1 Product Hunter while investing in AI startups I watch the floodwaters rise in real-time.

What s become clear: SaaS is dying; VC is withering . Building software is not uniquely compelling. Code has become a commodity.

What most people miss about commoditization is that when a product or resource becomes abundant, it doesn t just get cheaper. It unlocks new and previously uneconomic uses.

we built something as an afterthought, basically a debug tool we left exposed because we were lazy about hiding it, and it turned into one of the most used parts of the product. meanwhile a couple of things we spent weeks polishing barely get touched.

curious if this is universal. what's a feature that you almost didn't ship, or built half-heartedly, that ended up mattering way more than the stuff you were actually proud of building?

Hi Product Hunt,

While building an OpenAPI SDK generator, I realized that real-world API specifications are often much messier than the examples we see in documentation. I'm curious what's the biggest challenge you've encountered with OpenAPI? Circular references? Inconsistent naming? Missing schemas? I'd love to hear your experiences.

I have a bad habit of trying new tools even when the one I'm using is perfectly fine.

I'll see something new, think it looks better, move everything over, spend time learning it... and then end up back on the old app a couple of weeks later.

But then there are other apps I've complained about forever without replacing because switching feels like too much work.

I ran git log on my main project this morning instead of writing code.

First commit is 1 November 2023. Today is day 1,038. There are 550 commits in it, spread across 103 distinct days. 935 days have nothing at all.

I ve been thinking about what actually makes me uncomfortable about letting an AI agent take actions.

It may not be the size of the action.

It may be whether we can undo it.

I've spent time on the sourcing side of e-commerce (based in Hangzhou, working with electronics manufacturers), and I'm curious how founders on the buying side actually vet new suppliers before committing.

A few things I'd love to hear from this community:

• What due diligence do you do before trusting a new factory/supplier you've never worked with?

• Have you ever gotten burned by a supplier, and what red flags did you miss in hindsight?

• Do you prefer working through platforms (Alibaba, trade shows) or building direct relationships?

I've seen this from the supplier side, so I'm interested in the buyer's perspective what actually builds trust for you when sourcing internationally?