I only see fresh engineers with very little production outage experience reaching for the newest DX-focused stack to save 1 hour once per year while 10x-ing their risk.
An urgent notice from the Railway team
blog.railway.com/p/security-…
"Look I made your whole product in a day, and it cost $10k in credits - now what". AI is turning leaders who worry about productivity, in-person offices, and performance cultures into expensive laughing stocks. If you value impact you have no issues.
Congrats to OZ!
S&P Global has entered into an agreement to acquire @OpenZeppelin, the security standard for onchain finance. The transaction complements S&P Global’s risk assessment and ecosystem development capabilities in digital asset markets, enhancing its ability to create the next generation of onchain security assessments, benchmarks, and deliver essential intelligence as capital markets transition onchain.
Read the full announcement: okt.to/qcLufd
Honestly, even product factory may be too narrow - but this is the directionally right way to look at it. Perhaps it’s an outcome factory.
I came across the term "meat proxy" today: It's a person who takes questions from humans, proxies it to claude, and then proxies the response back to humans.
Steven retweeted
Virtual roundtable on August 12, 10am ET ⭕️
Market and counterparty risk now sit alongside smart contract risk, and few institutions have a framework for both.
Apply👇
us06web.zoom.us/webinar/regi…
en.wikipedia.org/wiki/Cargo_… This is my favorite historical metaphor for when practices are copied without being understood. Never blindly mimic and assume an outcome follows.
Development now belongs in CI. The real things we focus on include specifications, tradeoffs, and ensuring the needs of the business are met. The tools just need to catch up to this.
AI replacement of engineers affects non-technical folks, because the best engineers I’ve known are better at business than business folks. You’re about to have some difficult competition :)
Steven retweeted
Two years ago, Sei v2 arrived as a bet that the EVM could be the home for high-performance ecosystems.
One hundred million EVM wallets later, the call was right.
Next, Sei Giga will prove that the EVM can be the home for global finance.
One thing I’m learning is that AI tooling makes everyone feel like an expert at how to use it. Always prioritize learning from those who make a real impact. Often impact isn’t caused by their process/tools as much as values/priorities.
To improve node cold starts, v6.3 introduced:
- Sequential snapshot prefetch to replace random I/O
- Smart MADV hints to guide the kernel
- Selective preload of only the big trees (evm, bank, acc)
- Skipping the huge sparse KV file
- Skipping prefetch entirely if >80% of pages are already cached
This is a huge reliability and performance win for Sei's node operators.
Shoutout to the Labs engineering team.
You can view the changelogs for v6.3 here: github.com/sei-protocol/sei-…
AI is doing to code what the cloud did for infra. Just as we destroy and create environments, it will be easier to rewrite the whole codebase. Ephemeral code will change how everything is built.
A multi-proposer network offers interesting properties for MEV. As volume increases, extractable value will be key in keeping transactions affordable for all. Sei is laying the groundwork for the first sustainable low-cost network now.
This is why we do all our performance testing with varied levels of unique accounts including varying rates of new accounts.














