Connor McEwen

Fun announcement - more funding for more Fun

Having kids is a great coping mechanism for the AIpocalypse: - Too busy to have time for an existential crisis - Protect against AI work intensity burnout via childcare responsibilities - Easy way to assign purpose in life

Vitalik: The L1 is scaled and plenty of blockspace, don't worry. The L1 in question:

If you think of EVM chains (especially Ethereum and EVM L2s) as the JavaScript of the space - it becomes increasingly hard to justify another "language" (VM/chain) for app builders serious about distribution when those chains experience downtime at this stage of maturity.

@tarikbell @MorphoLabs @compoundfinance @Uniswap Big W for Morpho and Tarik

im thrilled to announce im now head of engineering @MorphoLabs 🦋 ive spent most the last 7 years in defi (fun fact - my first startup was built on @compoundfinance) and im as bullish on morpho as i am @Uniswap. why morpho is inevitable and what i’ll be focused on 🧵

This week I started at @FUN as CTO. Very excited to be working with @AlexKFine and the rest of the team - more soon 😁

The engineering team deserves a huge shoutout for making all this happen and putting their trust in me. I'm excited for everyone to see the things currently cooking make it to production, and proud to say that Hyperlane is key infrastructure in building a better financial system.

Alongside onchain protocol development (HWR 2.0), there's been immense growth in the offchain stack's reliability and feature set across the CLI, Warp UI, Relayer, Validator, and SDK.

Today marks my transition to an advisor role for the team working on @hyperlane. We're passing the torch to @trevorkporter as the new Head of Engineering.

Welcome @MicahhParsons11 to the @packers! Joining the historical list of huge defensive additions to early career franchise quarterbacks...

I can't say why the media angle on this stat is the headline but I can say that as someone leading a team, if you have found 0 ways to make your team more productive with ChatGPT/Claude/Gemini today, you are not trying at all.

The same data shows that 80% of teams who piloted a foundational model into their workflows considered it "successfully implemented". This matches the literature around people crushing custom models with foundational models + RL.

Offer my "guy not holding AI bags" take on the "only 5% of AI deployments are successful" stat: it's very misleading. If you look at the chart, it applies only to custom trained models.

I had a way better experience with Claude generating HTML for a document vs ChatGPT defaulting to PDF and having a bad time. Is this the end of PDFs when you can insta-convert them to a better markup language?

Early stage startups have to develop many processes over time. Previously, someone would pull the "let me see if I can get the doc from my old company that governed this process since it worked well". Multiple times recently I've gotten a new doc in 5 minutes with an LLM - wild

Nice writeup on how it's done x.com/_daniel_savu/s…

The "dark forest" of the mempool evokes a very adversarial environment, but there is also also an adversarial environment in the long tail of new chains and VMs. Simply indexing data and landing transactions reliably is a tall task