I spent the last week trying to answer precisely what makes a good eval. I trace the eval landscape from GPT-4 to today and tried to define the structure that good evals share. It would mean a lot if you could give it a read and share your thoughts :)
Link below!
Introducing OpenAI Guaranteed Capacity: a new offering that enables customers to guarantee long-term access to OpenAI compute.
We’ve made long-term investments in infrastructure, partnerships, and capacity planning to help customers scale reliably.
Now, Guaranteed Capacity helps customers plan ahead for critical workloads in a compute-constrained world.
https://t.co/TN4OkZr2Uo
the strait of hormuz is in a quantum superposition of open and closed that only collapses when you try to take a tanker through yourself and see if you get shot at
Just had the privilege of watching the film with the directors commentary experience for #ProjectHailMary
Here is a list of things I learned/found super interesting ! :D 🧵:
YC and Delve have parted ways.
I still remember the day we took our YC interview at MIT.
We’re so grateful to the community and every founder friend we’ve made.
We'll continue to support every young founder striving to make the world a better place.
It’s more so “what did the phones next to you search for”. If you stood in the vicinity of another phone that searched the blueberry store, chances you are also thinking about the same thing
It’s also frequency illusion. You recognize targeted ads more often than regular ads
Hey, Copilot, How can we get more 9s of reliability?
Copilot: *thinking*...
The user wants more 9s. Checking current uptime. 90.21%. Oh Dear. Only one 9 of reliability. The user is absolutely correct. I must get more 9s... Wow, wonderful. I've found the root of the issue and can offer an elegant solution...
Copilot: Let's reduce uptime to 89.99% to return to three 9s of availability.
Hii! I'm a recent Berkeley grad (math/ds) looking for a job.
While in college, I handrolled Fortran code working on a radiative transfer engine at NASA.
I'm now agentmaxxing and am very easy to get along with.
About me: I like Thomas Pynchon, Broken Social Scene, and Twin Peaks
When asked for a refund on an item sold in the vending machine (because it had expired), Claude promised to refund the customer. But then never did because “every dollar counts”. Here’s Claude’s reasoning.
$GOOGL besides Gemini has a 10-15% stake in Anthropic and a 7% stake in SpaceX that just bought xAI.
This means that $GOOGL now has exposure in 3 of the 4 frontier AI labs.