sari
@sari
making Sublime — for you and for me.
sari
twitter.com
But ambiguity, taste, meaning, value, these remain hard to attain within the hierarchy of verifiability. “Intelligence” races ahead wherever the rewards are liquid and the outcome is easy to check, but it lags at the edges: in murky context, slow feedback, or spaces where the answer is more felt than measured.
While it’s tempting to believe a model... See more
While it’s tempting to believe a model... See more
Market of words
Choosing your metrics, KPIs, etc...
listen for truths that language cannot name
But what felt new was the speed and violence with which language now manifests markets. I’d sit in meetings as a single phrase “AI-native vertical SaaS,” “verifiable compute infrastructure,” “narrative-market fit” jumped from one tweet thread to demo days to term sheets to LinkedIn profiles, mutating as it spread. Words weren’t just signals,... See more
Market of words
reading, watching & listening to
had to read this several times to grok it but applied to a real-life example.
It’s hard to design the perfect onboarding. but it’s easy to check: “Did the user reach ‘Aha!’ in 3 steps instead of 7?” or “Did 80% of users complete the flow?”
So the industry (and people) naturally get pulled into automated loops. this is why mid-training is powerful and dangerous. it can manufacture the appearance of discernment while dodging deeper unmeasurable criteria.
“They have learned to produce the appearance of discernment without ever developing the faculty itself.”
His authority is not intelligence but fidelity, the willingness to remain faithful to something that arrived without justification and refuses to submit to it. This is the thing no training regime can teach. Not taste, not speed, not range, not even the ability to reason, but the capacity to be remade by something unexplainable, and to let that... See more
Pre, Mid, Post-Training Way of Life
we build temples to ideas we no longer believe.
Scott Alexander’s recent post argues that the best response to AI displacement isn’t optimization or doomerism—it’s play. Figure out what you actually want. Do the thing you love doing.
This is harder than it sounds. We’ve been socialized to optimize for homogeneity, not edge. But just as retail orderflow now moves institutional markets, the... See more
This is harder than it sounds. We’ve been socialized to optimize for homogeneity, not edge. But just as retail orderflow now moves institutional markets, the... See more
2026 Open Inquiries
how do we get people to awaken into their power and do stuff?
Latvia’s Flow won the Oscar for Best Animated Feature and was created because he was somehow obsessed Blender, the free/open-source 3D modeling tool. It’s now consensus that it’s just a matter of time that an AI-made film could be a worthy contender in mainstream awards. Creative production pipelines are increasingly “node graph native.” ComfyUI’s... See more
2026 Open Inquiries
The Problem with Benchmarks
Benchmarks are often designed for papers, not products. The result? We get IFEval – a test that measures whether AI can make the letter "n" appear three times. Riveting.
Many datasets themselves are deeply flawed. That's why 30% of "Humanity's Last Exam" answers are wrong and 36% of HellaSwag contains errors. Garbage in,... See more
Benchmarks are often designed for papers, not products. The result? We get IFEval – a test that measures whether AI can make the letter "n" appear three times. Riveting.
Many datasets themselves are deeply flawed. That's why 30% of "Humanity's Last Exam" answers are wrong and 36% of HellaSwag contains errors. Garbage in,... See more
Benchmarks are broken. Here's why frontier labs treat them as PR.
superintelligence teams rely on human evaluations. humans can measure nuance, creativity, and wisdom – things benchmarks can’t.
