Building @cipherowl - ex-Coinbase, cruise, twtr, amazon and msft

Joined October 2008
208 Photos and videos
Pinned Tweet
Mar 31
Replying to @jack

1
3
10
6,372
leozc retweeted
May 16
thorchain gg20/tss attack path i reproduced the suspected gg20 leakage mechanics against the tss-lib version they used. it accepts malformed paillier material, exposes a type 5 / type 7 oracle shape, and the go-tss wrapper misses some important checks. banteg.xyz/posts/thorchain-t…
19
34
322
59,493
May 7
The most interesting line in crypto x AI is not “agents will trade tokens.” It’s that agents will eventually move money, buy compute, pay for data, settle with services, and act across financial rails. That only works if risk checks become programmable, explainable, and real time.
2
125
May 2
I agree next weave of software is dynamic software - like the google ads targeting … Building Pi, and what makes self-modifying software so fascinating youtu.be/n5f51gtuGHE?si=odrH… via @YouTube
1
4
109
leozc retweeted

39
81
343
50,121
leozc retweeted
Apr 23
Replying to @garrytan
The skill file is the new source code, and LLM is the new interpreter that powering intent-based computing.
1
3
147
Apr 21
This is pretty sophisticated but this is actually exactly what I have been mentioning: decentralization is not a choice, it needs to be by design. If your decentralization is not properly done, mixing centralized and decentralized elements, and your centralized point is what usually gets weakened and breaks faster than the others.
1
3
106
Apr 16
2 years ago
2
88
leozc retweeted
Mar 31
Replying to @jack

1
3
10
6,372
Mar 30
If you are a technical founder who is not currently writing code, you are throwing away 10-20 equivalent engineering headcounts of 2024.
4
101
Mar 29
Shipped ghostwriter -- autoresearch for blog posts. You write a draft. 8 AI agents score it from different expert perspectives (Claude codex, in parallel). A writer agent makes one edit targeting the weakest score. If the edit doesn't improve things, git revert. Loop forever. The draft never gets worse. It only gets sharper. More importantly, we keep all the artifacts from agents in each version - so much to learn! github.com/leozc/ghostwriter… Thank you, @karpathy, for such inspiration!
2
6
346
Mar 28
Onchain Investigators, if you are interested in helping us improve the tool, pls PM me and I will give you free lic and credits!
2
62
Mar 26
Yup!
SEC Chair Paul Atkins: 🇺🇸“All U.S. markets will be on chain within two years.”
108
Mar 26
Any payable SaaS needs to solve long term memory issue - that is where the value is
New on the Anthropic Engineering Blog: How we use a multi-agent harness to push Claude further in frontend design and long-running autonomous software engineering. Read more: anthropic.com/engineering/ha…
99
leozc retweeted
Mar 25
Replying to @bcherny
Excellent - thanks for letting us know and learn! Great ceiling line and which will be used to raise our baseline
2
1
4
2,789
Mar 25
8 new emoji, while my Siri is still retarded!
2
143
Mar 24
Onchain transactions will be 100x by 2030, but we have a hard time understanding all the details we should - luckily, we will have a very capable onchain investigator very soon!
2
8
595
Mar 24
Building an AI productivity application for an existing vertical is not just about using AI; it’s about fundamentally transforming what people do, how they think, and how they achieve efficient results — that’s the true challenge.
1
53