I used to only engineer software; then I learned you can engineer everything.
Building AI agents to automate supply chains.
Cofounder @ Elm AI | @Cornell CS
im convinced this is a sama psyop to make cost a big deal
oai is the uncontested king of cheap models that are very smart
they have been on the "intelligence too cheap to meter" train, and im 100% sure they have a clear line of sight to oom cheaper inference that anth may not have
if i knew that was my advantage, id talk about cost all day
everyone who codes knows claude limits are an absolute yoke compared to codex
OpenAI is absolutely cooked. This is loser language. You can’t be four years into the bubble saying “yeah our customers have a huge issue with how expensive our business is.” You just raised $122 billion! You can’t say shit like this!
https://t.co/1uKEEpSS03
i have a fun one, a mentor once referred us to a fund who asked us to pay (cold hard cash) to be able to pitch to them
i knew it was nonsense, but to keep in good books of said mentor (who is legitimately an incredible person, and has been super helpful in the past) decided to pay (for the story)
their process was total horse-radish:
- made to apply through a stupid, glitchy portal (v long)
- pay a fee (something like $200 ig)
- do so many meetings, it was like 5-10 hours of meetings over months (at one point, my co-founder and i sat in the meetings for morbid curiousity, it was astounding how they kept finding reasons to get on a call)
- they made us do a pitch event (along with other scapegoat founders)
- they wanted to talk to our leads in that round (we made the intro, again like morons); i just saw her email and she said she "finally had quorum to begin due diligence" whatever that meant
- all of this, over months, only to pass on us
- and to top it all of, one of the investors reached out to us to try to sell her consulting services
it was incredible, like performance art
I was once pitching in a board room at a top 3 VC firm for a $15M Series A.
12 people in the meeting. One of the GPs fully fell asleep. Out cold for 30+ minutes. Nobody acknowledged it. Everyone just kept going.
I kept presenting my Series A slides to an unconscious man in a Herman Miller chair and somehow that was considered normal. That's venture capital.
You might fly across the country to perform for people who may or may not be conscious.
It's a dance.
And sometimes you lead and sometimes you follow and sometimes your partner is unconscious.
If you're raising right now, just know: every founder has a story like this. The process is weird. The power dynamic is weird. You're not crazy for thinking it's weird.
No one talks about it because they want to continue raising. But I'm happy to stick my neck out there.
It is weird.
i have updated my priors on a kind of investor i had innate prejudice against -- the non founder investor
i have met enough really good non-founder investors to know that investing and starting a compnay can be decoupled to some degree (but not fully)
a good non founder investor tends to be excellent at pattern matching, and info absorbing *while* being zero ego, and high humility
the second characteristic (humility) has done the heavy lifting in all my interactions, converting these folks into true value adds -- they come with their insights but immediately know when they don't know about a certain topic since they haven't ever built a company
there are some non builder investors who do not have humility and confidently/aggressively give second hand advice ("teaching you to be a world class swimmer, without having stepped into the pool") that spoil the reputation of other investors.
thankful to all the good non-founder investors i have had the good fortune to interact with
just landed in NYC.
what’s everyone’s must-get food when they’re in town?
mine is usually whatever the newest / hottest bagel shop is
but i need more ideas
side note: $150 uber into manhattan from the airport is completely insane
@gregpr07 zed is one of the few modern pieces of software that feels genuinely well crafted
I used zed for a few days, and now every other app feels slow and annoying to use
imo the best move here if budgets need to be slashed is to give them access to cheap models that are still good (oss models) and force the engineer to learn how to deal with agents better
the newer models are smart enough to get intent, but cheap models can get similar outputs if the swe is skilled enough
I actually feel sorry for Uber
$1500 per month can be very low if your engineers have gotten used to stupidity like asking Opus 4.7 to do git commit.
But then again, "token efficiency" workstreams will create a lot of promo packets, as it happens in bigtech all the time.
Excited to announce Slashy
The first email client that works for you.
The real cost of email isn't the time. It's the mental load of constantly checking it, just in case something needs you.
Slashy kills that. You never need to open your inbox unless Slashy tells you.
Try it out at https://t.co/6tuQxWV7KR
she brings up a good point at large, it doesn't matter whether you used llms or not, intent/thought/care in your work matters
LLMs are often made a scape goat for poor quality output, when in fact lack of care/thought from the author is the problem
For corporate america = care in creating deck matters
For founders = care in creating product that users love matters
For software engineers = care in building a clean, well thought code base matters
LLMs don't change these things. If LLMs have changed these things, it implies there was no thought in the first place.
@AdvaitRaykar The difference is in corporate America and most of the world it matters
People want to feel special if they’re buying something from you, not that you put 0 thought into generating the slides
I would say here that putting 0 thought in the slides doesn't necessarily equate to using LLMs
Same applies for code, llms' can produce stellar code and llms' can produce pure slop
context, references, and good prompting are key
Some of the best (and worst) decks (and code) I have seen have come from LLMs
I think LLMs should not be made a scape goat for the lack of thought/intent from the author using the LLMs