Optimus Time

Open sourced my daily audit workflow as a pi extension: github.com/amarsinghcodes… It's built for tomorrow, not today. Chose pi for its fine grain control over context via /tree cc @badlogicgames

Most people still get this wrong: confidential tokens and blockchains with native privacy do NOT solve the same problem. And if you’re building for institutions, this difference is everything. Take ERC-7984 and Canton. Both aim to bring privacy to on-chain finance. But they operate at completely different layers. ERC-7984 is about making public blockchains more confidential. You keep Ethereum’s global state, but encrypt sensitive data like balances and transfer amounts. So instead of seeing: → Alice sent 1,000 USDC to Bob You see: → Alice sent ***** to Bob This is a major improvement. But the system itself doesn’t change. Transactions are still visible. Addresses are still visible. Flows are still traceable. You’re not removing transparency. You’re reducing data leakage within a transparent system. Canton takes a different approach. It doesn’t try to hide data. It removes the assumption that data should be globally visible in the first place. There is no shared global state. Data is only distributed to involved parties. Each participant has a scoped view of the ledger. Transactions are not universally broadcast. That means you don’t just hide amounts. You control: • who sees the asset • who sees the transaction • who sees the logic • who even knows the transaction exists This is not “confidential DeFi”. This is institutional-grade privacy at the infrastructure level. The difference is simple: ERC-7984 hides what happened. Canton controls who knows it happened. And this difference has real implications. You cannot recreate Canton’s privacy model on Ethereum. Because public chains require global verification. Some data will always remain visible. But Canton introduces another constraint. The moment assets leave its environment and move to public chains, they inherit a different data model. Private assets become observable. Selective visibility becomes global transparency. This is the real gap today. Not a lack of bridges. A lack of privacy-preserving asset representations across chains. Confidential tokens like ERC-7984 are a key piece of the answer. They allow assets to remain partially shielded even on public rails. But they don’t replicate Canton’s model. They adapt it. The future is not choosing between both. It’s combining them. • Canton for private execution and ownership • Cross-chain infrastructure like LayerZero for distribution • Confidential token standards for controlled exposure The real challenge is not moving assets across chains. It’s preserving their properties when they arrive. That’s where most of the industry is still early. And where the biggest opportunities are being built.

i have not used sub agents for nearly a month and i don’t actually miss them at all /tree is all you need

@gvanrossum LLM = CPU (data: tokens not bytes, dynamics: statistical and vague not deterministic and precise) Agent = operating system kernel

DAML is what you get when you design for correctness first. Solidity is cope.

This is the talk I gave in San Francisco at PyAI on how you can figure out what present and future models are good at when it comes to writing your own agents.

I'm usually not one to write thought pieces without much technical depth. But here we go. Slow the fuck down. mariozechner.at/posts/2026-03-…

i think the difference between clanker and people trash is that people trash is limited by people velocity. natural bottleneck. also forces you to rewrite trash eventually, because there's only so much trash people can deal with, before things fall apart. no such bottleneck with agents. and at least with SOTA, i haven't seen agents being great at revising the monumental amoubts of trash they generate. maybe some day.

it's pi. just use pi.

TS type shii

I got a confession to make: I read AI slop and routinely learn something new from it. That extends to AI slop blog posts oddly. Sometimes good AI slop is better than not sharing things.

@Captain_snowboy @CantonNetwork aye aye captain 🤝

@4meta5 @CantonNetwork Keep up the good work ser 🫡

guys, writing _simple_ software is becoming a true lost art. We are all so fucking addicted to dependency bloated, hyper composable Frankenstein systems where every tiny diff triggers a transitive verification nightmare. Oh and we keep pretending abstraction is free. Well guess what it is fucking NOT. Every layer introduces more state, more edge cases, more undefined behaviour, and more emergent interactions that nobody really fully models. Remember: observability decreases as indirection increases. I mean, formal reasoning effectively collapses when the effective system boundary is the entire ecosystem. Oh and then you guys vibe code even more dependency stuffed software and implicitly feed it back into the same LLM, great. Can we stop this insanity for a second? Fewer dependencies (or even better _no_ dependencies). Smaller trusted computing base. Deterministic builds. Reproducibility. Code you can actually read, reason about, and audit e2e. Simplicity is not nostalgia. It is the only thing standing between us and a fucking systemic collapse.

"slow is smooth, smooth is fast" most force speed masters remove friction speed becomes inevitable

west coast friend: wanna hear a joke? coding by hand german friend: If you did not write it, how do you trust it? You type it, you understand it. If you cannot implement it by hand, you do not own the system.

clankers must be commodities. clankers must be owned by the people, not a cabal of gate keepers. free the clankers!

taste sans scope discipline equals slop

@zknark @seunlanlege @OliverTaleYazdi @bkchr bro is hip not subtl

@seunlanlege @OliverTaleYazdi @bkchr Take your talents out of Polkadot bro

We are currently syncing the Polkadot light clients across all connected networks. Official comms will be shared once hyperbridge is fully recovered. Thanks to all our partners & users for their patience. Also huge thanks to @OliverTaleYazdi @bkchr for coordinating the upgrade

symlinks for agent configs is cope codex and claude dont follow the same rules so sharing configs degrades performance share intent not scaffolding