Ryan Hartman

This is a great read about a lovely future! I wish there were more ideas of good outcomes available for us to think about like this

I hope you enjoy the story! It's a short read, and I put a lot of love into it. corbt.com/posts/the-mach…

Very excited for this release! LLMs from Scratch was a beautiful dive into smaller LLM architectures and I expect no less from this

when you want to work with the smartest people in the world and you get the opportunity to talk to them but now you are nervous

Given the balancing act they have to do between inference and training, I wouldn’t be surprised if they had already planned to reallocate some of their Fable inference capacity to training or Sonnet 5 inference after June 23rd. USG pausing Fable for 2 weeks probably didn’t materially affect their plans to continue training better models on their original schedule, but the timing is likely affecting their allocation now that they are rolling Fable back out to everyone. High usage costs can reduce demand for a product, so this might be a band-aid step to try to maintain that balance. I’m pleasantly surprised that Ant offered to roll Fable back out to subscriptions at all. There are a ton of implicit incentives to stagger the rollout and only provide Fable to enterprise customers who are willing to pay significantly higher margins

I don’t understand Anthropic here. Reports make it sound like they’re doing great financially, but bringing Fable 5 back for subscriptions only for a very short window, including it for only up to 50% of weekly usage limits, then pushing it into usage credits feels like aggressive monetization under pressure after this huge upheaval. We were originally going to get 100% usage from June 9th to June 23rd for 15 days. And why make the situation even worse right after a customer trust crisis especially when some sessions will route back to Opus 4.8 anyway? Maybe there’s more to it, but from the outside it just makes this whole thing worse. This is no hate against Anthropic I'm just trying to understand executive level decisions here. Am I the only one confused by the decision making here?

@beffjezos It’s a bit more like this

This is how every human brain has worked since the dawn of civilization btw

One of the key tenets of media theory says as much! If you’re interested in reading more about it, one of the first things that they talk about at Meta NHO is McCluhan’s book “Understanding Media: The Extensions of Man” which digs into the idea that the medium shapes much more of human perception/society/culture than the content within any individual message

Yes, YES!! This is what I've been trying to say!! The medium is the message: the format through which something is said content is the determinant factor into what is said — NOT the content!!

One of my mentors told me to leave Data Science back in 2024 when both of us were laid off. He took his own advice, but struggled to find work for months and eventually took a contract role as an IC product manager. He is now doing very well, but the journey was long and arduous. Lots of people in DS industry see the writing on the wall and would love to do the same but don’t have the fortitude to weather the transition. It can be hard to give up a high salary just because you have a hunch it might dry up one day

Has anyone ever actually recovered from an analytics career? Being a career analytics guy is like being an investing generalist. You can kinda do a lot of shit but it's all basically worthless.

What this means is the typical customer gets further from the frontier at exactly the same time that access to the frontier becomes economically valuable

Concretely, this means you could be coding innocently and suddenly run into lockouts on the parts of your product that are the most valuable. Consumers of Fable would be left with a model that is deeply inconsistent when it comes to building products.

Ant will be fine even if they don’t recoup the training costs from Fable, but the USG is effectively setting the precedent that everyone else will be relegated to the underclass once AI becomes sufficiently powerful

In America, a stranger will rename you in a single breath, and you are simply expected to come when called. I went to eat at a busy restaurant. A young man at the front asked for my name, to mark my place in line. I gave it the weight it has carried for eight hundred years. "Nobunaga." He smiled, nodded, and wrote it down with great confidence. Then he read it back to me, to be sure he had honored it correctly. "Perfect. Banana, party of one." Banana. He had heard my name, held it a moment, and returned to me something rounder and more cheerful. To refuse the name a host gives is to refuse his welcome. I bowed. I was Banana now. Then he handed me a small black disc, said it would "light up and buzz" when my table was ready, and turned to the next guest as though he had not just placed a living thing in my hands. I held it in both palms, the way one holds a small sleeping beast that may wake. I found a place to stand. I waited, ready. It woke. It screamed. It flashed red. It leapt and shook in my hands like a captured spirit demanding release. A lesser man would have dropped it. I did not. I gripped it, steady, looked into its blinking lights, and told it, in a low voice, that its time had come. Then I carried it back to the host with both hands, the way one returns a hawk to its master. He took it without looking and shouted across the entire room. "BANANA! Party of one, your table's ready!" A hundred strangers turned. I rose. I crossed that floor as Banana, spine straight, chin level, a man answering to his name. A child pointed at me. I gave the child a small bow. He had recognized me. All through the meal they kept me. "How's it tasting, Banana?" "More water, Banana?" The check, when it came, said Banana, and thanked me for visiting. By the end the whole staff knew me. They waved as I left. "Night, Banana!" So tell me honestly. For eight hundred years my clan answered to one name. Tonight I answered to a fruit, calmed a screaming relic in my bare hands, and ate among people who were glad I came. When the little disc lights up, is the table truly mine, or am I only keeping it warm for the next Banana? Because I have already decided to return on Friday, and to ask, very humbly, for the same disc.

Some updated standings: Opus 4.7 still on top, GPT 5.5 getting much closer. I am excited to see what happens as OpenAI continues to RL this supposedly new pre-train.

GPT 5.5 is catching up to Opus on the KahneBench! It has a unique cognitive fingerprint that I haven't seen from previous models in the GPT line, most notably, it is as a loss averse as the Claude models typically are. It also has a tough time handling base rate neglect biases, something that has been "solved" for most models released in the last 3 months