If you think about "price", "speed" and "accuracy (reliability/quality)", our bet is that models won't hit those 3 together. So you won't get a model that is very fast, very cheap and very accurate anytime soon.
Also, imagine that you have a case where you want to scrape 10,000 records from a website, why have AI navigate to every page to do this? why not write the code, run it, and get consistent and fast result? its also predictable, if it messes up, you know what happened and you can trace it to the exact line of code.
I’d like to see less focus on Playwright and more focus on giving the agent more than just an MCP to browser automation. Make it multi-modal, figure out how to optimize when to send screenshots to which model, etc… current coding harnesses are awful at any UI automation because they’re just automating DevTools and occasionally screenshotting. It’s obviously robotic, it’s slow, it’s ineffective and makes it difficult for the agent to validate success of code changes.
Generalized computer use is what will ultimately solve this, but I think there’s real intermediate value in optimizing browser workflows specifically, as a medley of remote browser automation and multi-modal browser use.
our product is infra + agent. You can use codex or other agent to generate the code. We actually have a cli that allows you to deploy projects to our infrastructure.
We are actually working on open sourcing a plugin that you can use with any coding harness!
Fair. The differentiator is the Rust single binary +
petgraph knowledge graph. No Python runtime, no cloud,
survives restarts. Built it because nothing local fit
that profile.
how to capture, what to capture, when to capture it.. where to put it.. how to make it 'useful'.. how to reinject it or make it accessible
the harness makers may well come up with better means than flat files, but there are loads of folks out there working across different harnesses and in teams, and there's very little that works in that respect.
Cheers — quite a bit more work going into that presently and extensive testing to back it up. Early finding: auto-extracted memory is no better than a hand-written doc. Human curation is the lever. Numbers at the top of the mori README.
Let me rephrase: vibe-coding something that will solve 80% of a company's problems in this domain still takes a tremendous amount of work, more than it seems on first glance, and it isn't worth the effort for many teams. We focus on solving this problem to the Nth level, full time, so that a company can exclusively focus on the things that are truly specific to them.
I wonder if this approach to starting a vertical business the founders have 0 experience in has ever panned out. I know YC pushes the B2B SaaS angle as hard as possible, searching for "underserved" niches, but seems like if you don't have true industry experience, it can't possibly work out.
Travis Kalanick had zero taxi experience and Brian Chesky had zero hospitality experience.
Now they created new models to existing paradigms, because I do tend to agree that founders that have verticalized experience tend to be far more successful (but perhaps arguably less 'disruptive')
Those were consumer apps, not B2B. No deep niche experience needed, Uber had to fight regulations but wasn't something industry knowledge would've helped with a ton
it’s a good point. Their product likely is legit, but i even judge individual contributors who use these site builders for their resumes or portfolios and stick with the subdomain. It’s that last mile effort towards completeness i think.
Why didn't you found yourself at any point? I understand gaining experience as a founding engineer, but after a certain point, you need to work as hard as the technical founder, take the same career risk, for way less payout
I founded 2 of the companies above (1 acquihire, 1 sold big portion of my shares back to investors when I left).
Yes I agree founding is much better than being first engineer. Reason I kept joining as first engineer is because I found product market fit for my self (it's funny to say that haha). I was able to pitch myself as first engineer that builds first product and first teams. Someone super technical who likes to grow code and grow people is actually pretty rare in startup world so I was able to ask for high comp and choose to work on interesting projects.
I would define that as success. My definition is the feeling of success in the eyes of the founders, not millions of dollars or whatever hyped up benchmark
were your projects in spaces you were familiar with? I feel like that makes the dart less random. Also I feel like it's hard to get customers/users on calls
well they were all websites so they had that in common but not much else. I would recommend starting off with an app store of some kind since getting initial customers is a real pain unless you already have some lined up due to past work experience / networking. That said, with enough time and screaming into the void you will get someone to walk through the door. Best to ask them as many questions as you can while they are hanging around!
you can email me if you want: jason[at]zigpoll.com
Ideas came from building something that I wished existed. Honestly I didn't overthink any of them and I'm glad I didn't. If you do too much market research it tends to make your vision seem less exciting IMO.
Either way my whole point is to just iterate on your initial idea for long enough that it goes from being "your fun idea" to "your fun idea that people are willing to pay for". Then just repeat that loop until you retire :)
Yeah, I worked for a company in a relationship/middle man space that solved a major discovery problem at the time.
Due to some perverse incentives in the industry as well as some protectionism and legal chilling none of the players actually solved their customers fundemental problem. They got stuck as these weird tweener lifestyle companies.
I quit, moved to a cheap studio apartment in Brasil for 3 months (we still had a bully in the space put together a bs packet and used their connections to get some FBI agents to show up on our doorstep only to drop everything after our little interview), coded out a marketplace play and launched it.
Ycombinator flew us out but declined and a group in a smaller market gave us the only money we ever had to raise. Due to the fact that we built a legitimately fun marketplace play that solved a lot of frustrations in the space Scott Hanselman tweeted us out to his followers and that was all we needed to get the exposure and google juice ball rolling.
Due to the industries continuing perverse incentives, I could start the same company again tomorrow except it would take a big ad words dump of cash and I don't want to pull out of my own pocket yet again (I did just prior to covid but it's an industry that was totally shut down during covid and I no longer have the appetite.)
So by overnight I mean, we started generating huge buzz in the industry and 10x traffic of established players overnight and middlemen don't have motes so we never really had to raise, it was just the time to code out the platform and get it populated with data at the MVP level.
reply