Goldee Udani

Hacker, Entrepreneur, Philomath, Non-conformist, Interested in Psychology, Astronomy, Behavioural Economics, Business and Technology. Views my own.

[Goldee Udani @goldee]

When you ask an LLM to be critical, it can produce performative criticism, sort of plausible sounding objections that satisfy the user, which is just sycophancy wearing a different mask

[Goldee Udani @goldee]

Deep domain priors and confirmation bias are indistinguishable from the inside

[Goldee Udani @goldee]

AI is rapidly collapsing the software production costs to zero. High-margin painkillers are now massive targets for disruption - customers will swap them out for cheaper AI alternatives. The unexpected survivors? 'Vitamins' that fly under the radar due to pure inertia

[Startup Archive @StartupArchive_]

Keith Rabois on how to identify great talent

“What you want to do with every single employee every single day is expand the scope of their responsibilities until it breaks… and that’s the role they should stay in.”

Keith tells the story of giving an intern the task of getting smoothies to arrive at the office at 9pm to reward Square’s engineering team. And once the intern proved they could solve this task that had stumped multiple people at the company, he gave them something more important and consequential to do.

Everybody has some level of complexity that they can handle, and you want to keep expanding their responsibility until you see where it breaks—as Keith points out, some people will surprise you:

“There will be some people who you don’t expect—with different backgrounds, without a lot of experience—who can just handle enormously complicated tasks. So keep testing that and pushing the envelope.”

Keith also argues that you should monitor who is going up to other people's desks. If you see people frequently going up to a person's desk, it's a sign that that person can help them. Promote these people and give them more responsibility as fast as you can.

Video source: @ycombinator (2014)

[Guillermo Rauch @rauchg]

Obsession is the mother of invention

[Adam Grant @AdamMGrant]

The best way to gain job security is not being the most knowledgeable person in the room. It's being the most reliable person in the room.

Information is abundant. Dependability, helpfulness, and responsiveness are relatively scarce.

We count on people who consistently deliver.

[Goldee Udani @goldee]

RL + LLMs are unlocking ideas that were impossible before. This wave will disrupt more than every past tech wave—combined.

[Benjamin Todd @ben_j_todd]

Why can AIs code for 1h but not 10h?

A simple explanation: if there's a 10% chance of error per 10min step (say), the success rate is:

1h: 53%
4h: 8%
10h: 0.002%

@tobyordoxford has tested this 'constant error rate' theory and shown it's a good fit for the data

chance of success declines exponentially

[Dalton Caldwell @daltonc]

There is no substitute for a strong internal drive to do things, everything else is built on this foundation.

[Paras Chopra @paraschopra]

Hire high-agency people. Give them clear goals. Critically question their assumptions every now and then.

That’s it.

This is all you need to build a high-performance org.

[Shreyas Doshi @shreyas]

Outcome-focus can only emerge from genuine confidence.

This is why you will find outcome-focus missing among those who project bravado but lack true inner confidence. They will exhibit authority-focus, systems-focus, expectations-focus, outputs-focus, but never outcome-focus.

[Justin Halford @Justin_Halford_]

“Self play allows you to turn compute into data.”

- Ilya Sutskever, 2018

Just after this portion of the lecture, he discusses agents. It is clear that autonomy is the next step.

Life 2.0 will be a compute-bound phenomenon. ASML and TSMC are the midwives.

[Goldee Udani @goldee]

@amnigos @khoslaventures @z47_vc @BatteryVentures @BlumeVentures @peakxvpartners @neon_fund @stormventures @atomicworkhq Congrats, @amnigos !

[Hiten Shah @hnshah]

Most teams think execution is about writing code. But it's really about making decisions. The best engineers aren't the fastest coders, they're the clearest thinkers.

[Garry Tan @garrytan]

Great founders are concise and direct communicators. They are not in pitch mode. They use simple declarative sentences.

The word itself becomes reality when you can be that clear. @bradflora describes one such YC startup 👇

[Garry Tan @garrytan]

Clear logical step by step thinking is the antidote for hallucination and self-delusion

… in machines and humans both

[Nassim Nicholas Taleb @nntaleb]

FOOLED BY RANDOMNESS
My conjecture, expressed on X, that pple w/news in advance don't do well has been tested by Haghani et al.

Indeed they failed to really capitalize on information 1) you don't know beforehand what is noise, & 2) overestimate the information (sizing)

[Gaurav Sharma @GauravSharma]

"The market is like a large movie theatre with a small door, and the best way to detect a sucker is to see if his focus is on the size of the theatre rather than that of the door" -@nntaleb

[Goldee Udani @goldee]

@grease_ @businesstoday @sprintoHQ Congrats, @grease_ and @notraghu!

[David Lieb @dflieb]

I’ve made or seen hundreds of hires, and one thing I have never seen is a B player hire an A player. Not once.

If you play that out, you realize that compromising on even a single role can destroy a team forever.