Stumbled on this ancient (2025) resource on training "calibrated" classifiers fireworks.ai/blog/fine-tune-…
"Model routers require reasoning"
"LLM-as-judge requires reasoning"
*taps the book*
Any system 2 (slow, methodical) skill can be baked into system 1 (instinctual) with practice
Better yet, a pattern-matcher can detect a non-match and thus escalate it to a powerful model
We rotated keys for ~20 projects and moved a great deal to Infisical after the Railway meltdown. Was quite a slog but likely worth it.
As we say in Toronto, "collect your mans"
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios.
The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios.
Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said.
Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior.
Read my latest for Axios here: axios.com/2026/10/22/ai-agen…
The coding agent’s preferred unit is moving up in scope:
- the line of code (early Cursor)
- the file
- the branch (early Claude Code)
- the repo
- the machine (GrokBot, Dots)
- the internet (whatever tf is going on right now)
Something new I've been working on @RubricLabs: experiments
In my view, statements of "I think X will Y" can increasingly be answered by a small vibe-coded experiment, so long as the result is replicable
For example: can Jev mask PII? Turns out it can!
A disclaimer: the write-ups are straight ClaudeSlop for now, optimized for getting quick, directional answers
Another one: does enabling OpenAI's native (encrypted) compaction improve coding performance in Pi?
Probably yes!
You can see and expect more here; we fire them off as questions come up:
rubriclabs.com/lab
To win public support, AI needs to flex some deflationary power
If tokens get cheaper by 40x/year, at what point do high-quality foods get cheaper by even 10%/year?
Every part of the world now produces 2.5k+ cal/person/day - surely this is solvable
Seeing more benchmarks using $/decision or $/task
Seems like meaningful progress toward outcome-based pricing / incentive alignment with the user
Ted Spare retweeted
⌘⌥✗⚭✲≠
Meet Common. Every symbol your keyboard is missing, built for the Mac.
No login. No subscription. Free forever.
Available at common.wtf
Ted Spare retweeted
On the Rubric blog you can actually try out these demos yourself!
We think a lot about how to share ideas as accessibly as possible without wasting the attention of our technical audience
We want our blog to be easy to skim and handle deep attention if you decide to pull over
What can you build with Jev?
Practical use cases for fast, cheap decision models.
rubriclabs.com/blog/what-can…
So many activities!
What can you build with Jev?
Practical use cases for fast, cheap decision models.
rubriclabs.com/blog/what-can…
Ted Spare retweeted
What can you build with Jev?
Practical use cases for fast, cheap decision models.
rubriclabs.com/blog/what-can…