Founder & CEO for @santorilabs. Ex-GM @scale_AI (human data for frontier models), ex-@Google. Writing about how AI changes how we work and think.
San Francisco, CA
Joined April 2013
- Tweets322
- Following190
- Followers216
- Likes277
questions for designers: how do u do design thinking when the real users for your products are the agents?
1. empathize how?
20
We’ve spent billions rewarding machines and forgot about humans.
So, what is the real reward function for humans?
The process of articulating a problem is often more valuable than the solution itself.
Schmidhuber argues that curiosity is the satisfaction of discovering patterns that make seemingly confusing things become understandable.
Imagine a difficult engineering problem. You spend all night dissecting and struggling with it.
Then, you finally solve it.
The interesting part is that the problem always stayed the same. The only thing that changed was your understanding.
Maybe that’s part of the reward function. You learn best from friction.
2
106
working at google and scale ai, to now building santori labs, you'd think I'd seen enough chaos
yet none of that could've prepared me for what I witnessed yesterday on X
6
382
Back when I was at Scale AI, I was working on a Mandarin data project.
As a native speaker, this should have been one of the easiest things for me.
Turns out, it wasn’t.
I knew my writing wasn’t good and I couldn’t figure out how to make it better.
Oddly enough, this is how I build products today.
I’m good at knowing when something looks bad. If I’m unsure how to fix it, I have AI generate options and I pick.
Sometimes I stare for hours but the judgment is still mine.
That’s the contrast that matters.
7
387
everyone knows this and 99.9% of people takes it for granted. meanwhile, we take it a personal goal to eliminate meetings from our company as much as possible
there are way better ways to collaborate now, if you enjoy a meeting, have it. But so many of them are waste of time and mental health
Imagine a study showing that meetings cost companies more than AI tokens. A $200k employee spending 4 hours a day in meetings is basically $100k a year of meeting time. Put 10 people in the room and that’s almost $1,000 an hour. So many meetings have no agenda and so many people invited when only 2 or 3 actually talk.
1
1
158
I hate sf halloween because what do you mean you’re dressing up as an anthropic safety engineer
1
3
161
one of the best decisions a company can make in a rapid-moving time like this
maintaining or trying to grow a meh product comes with insane opportunity cost, especially for oai
OpenAI kills a lot of its products
Sora lasted 6 months, Agent Builder 8 months, Group Chat 8 months, Atlas 9 months...
Most of the time (Sora excepted?) their own product became obsolete within months, in case you think they are immune to misjudging the exponential themselves
3
280
"all path in tech leads to human data"
if you’re an employee of any corporation, in the medium & long term you are effectively generating training data for an ai that will likely replace you.
even if the company fails, your work will still create value as training data.
i know a lot of ppl find this bleak & possibly unpleasant but it’s the truth. & it’s kinda cool to think about. humans were never meant to do jobs that could be automated (we don’t use plow’s anymore). eventually we should all be creating our ai replacements in some sense so that we could live the way we were meant to.
4
249
lots of good points here. @TheRealFlups and I have essentially ask our agents a lot on - is this something you two are already working on
An good ai teammate should absolutely take the learnings and apply across the board. Imagine talking to your hrbp or manager who has zero context on what's going on with other people/teams
2
144
here is the thing: your best people are motivated when they are surrounded by great people, in an env where they can thrive.
The rest are either empire builders or slackers. I think we still need a few empire builders. but that's it
Just spoke to the Head of AI at a huge company.
The #1 thing they're struggling with is... what do I do with the AI laggards at our company?
Like most companies, they’ve found that a Super Pareto Principle is true within their org. 10% of their employees are responsible for 90% of the uplift in productivity/token usage from software engineering to non-technical knowledge work.
They’ve spent a lot of time & resources on AI enablement from LLM foundations to function-specific trainings and skill building. And yet, the power law of productivity increases hasn’t budged.
So now they’re left with a few sobering thoughts and hard questions:
1) 90% of people are overpaid and 10% are underpaid
2) We’d be better off allocating tokens proportionally to the power law, but how do we do that without killing culture?
3) Significant layoffs are likely warranted and wouldn’t impact overall output, but cultural backlash is a big concern?
4) How do we hire more power law people as a non technology company?
5) Is all of our efforts on training and enablement ACTUALLY worth it?
6) How do we not experience a brain drain where the top 10% all leave because they want to be surrounded by other AI-native 10x’ers?
2
340