BinRoo

> Harnesses go from internal tooling you’d happily buy to something you’d no more outsource than your product-eng org or your GTM team.

I couldn't have said this better myself. If you're not owning your harness, and you're not owning your model, then that leaves very little moat for any AI native company.

For anyone who can't make it to IEEE Quantum Week this year in Toronto, I've made my accepted tutorial publicly available. New lessons drop every Tuesday.

Beware of the benchmarks listed. SciCode and EnterpriseOps for instance: https://shukla.io/blog/2026-08/gym.html

Since the donut is already leaning "entertainment", make it bipedal and I'm sold!

Aside: the AI summary of this attributes the wrong person. This is Andrew Wiles.

Love that this goes beyond paged attention. Curious how this compares with Radix Attention [1]?

[1] https://sgl-project-sglang-93.mintlify.app/concepts/radix-at...

Food for thought: if inflation remains poor (3.5%) a $5 hot dog will cost twenty-fold in under a century.

I ventured out to better read Qiskit code, and came back with this fun idea of 2D expressions

> In the AI era...

Back in my day, you could also just Google the problems and find the solutions. What mitigated cheating at UVA was the honor code and each professor's faith and trust in our integrity. That culture was enough to not cheat.

Imo, the fix should be to work on culture. Cheating should always be a tempting choice, so that the student may challenge their integrity, which is a muscle that can atrophy.

Beautiful write-up, thank you.

One aspect that still bothers me is that you claim the just-say-no-engineer "was a critical role during ZIRP." I might be in the minority here, but I don't hold that same stance. I wonder if I am alone in that?

Ever read a quantum circuit? It's all so foreign, so let me introduce to you a simple one (no speedup) called rank-select, described through an interactive story.

Great ideas, though not a huge fan of pre-commit hooks that run full CI locally :)

I'd like to add another idea: automatic PR merge contingent on another PR getting merged.

Python DSL for safe quantum programming that compiles to Qiskit, where the type system enforces coherence and ancilla cleanliness

https://github.com/binroot/b01t

LLMs fall victim to "garbage in, garbage out." Claude can solve open problems if you know what you're doing, but it can also incorrectly convince you it's right if you don't know what you're doing.

A PhD teaches you how to think, how to learn, and how to question the world. That's a vital set of skills no matter what tool exists.

Rousseau said "mathematical precision has no place in moral calculations," so I was tempted to see how far I could go. Started it during the holidays, but finally came around to putting a bow on it :)

The human still needs to think, of course. But, I can get to my answer or my primary source using a tool faster than a typical search engine. That's a super power, when used right!