@xlr8harder my hunch is Gemini due to long context capabilities

@IterIntellectus > casually giving your watch to someone who compliments your watch

@crypt0x_0 I keep alternating between go and zig for my next project but honestly it’s been hard to decide which one

Monetization is about exchanging money for value. Monetization should not be about losing credibility for money.

@KevinNaughtonJr You forgot the hardest part: > naming data

@AlexHormozi > meetings that don’t seem like meetings

@a__demiroglu @AravSrinivas I think this can can be something that’s heuristics based using an offline eval of different query types and seeing what model performs best across different categories. This can also help bootstrap a custom router down the line.

@DeryaTR_ wondering if he decided to pivot after seeing R1 so it will take a while before we see anything

@nosilverv looks like something USAID funded

it’s easier to move bits than move atoms. This means people who move bits have had more development/iteration cycles than people who move atoms. This means they are more aware of common pitfalls and optimization strategies for any given problem. Taking people who move bits and having them optimize systems that move atoms works well because knowledge is transferable at the right level of abstraction.

“yo that was trippy” > no the universe pulled the curtain back to reveal something significant and you should reflect on it and make meaningful changes to your life as a sign of respect.

@tsarnick so deep seek is a model that does what the user wants. wonder why it became so popular

@Marks_Tech the Apple one actually bent space and time to disappear the object so it looks weird

Actual AI policy would look something like: > Fast track applications for data centers and power plants. > Clear legal framework to interpret copyright law for AI model usage. > Expand O1 visa to recognize meaningful technical/research contributions to the field with a path to permanent residency. > Create a public GPU cluster where anyone can submit training/inference workloads. > DARPA Grand Challenges around AI tasks with bounty. cc @DavidSacks @sriramk

Once upon a time AI research simply released stuff without safety concerns and the community (correctly) pointed out the issues with biases in the training data/model. The pendulum has swung too far to the other side now. Researchers/companies are too paranoid about even the appearance of potentially unsafe content that it is hurting progress.

@whstancil yes the correct way to do this is to solicit RFP and wait 6 months to spend 100x the cost of goods

@paulg @elonmusk systems evolve or they die

@briantylercohen did you file taxes using this service?

@elonmusk I’m surprised that the bond markets have not priced this in.