0xProtoCon

Protocol Builder | AI Power User | My views are my own

[0xProtoCon @0xProtoCon]

I'm excited to launch something I've wanted to build for years

Unveiling my latest cook: CLOUT CARDS

Everyone wonders when crypto apps that people want to use are going to show themselves. I got tired of waiting, so I built something I'd want to use every day

I picked something that was traditionally hard to do onchain, and difficult to do right with trusted computation: Crypto Poker

Clout Cards is gasless, verifiably fair, socially integrated to X, and realtime

No more high gas costs, no more shady databases, no more botted tables

Check out our testnet alpha at CLOUTCARDS (dot) FUN

[0xProtoCon @0xProtoCon]

@McGreenBeats Paging @Rengle820 😂

[0xProtoCon @0xProtoCon]

@kai_xbt Or he keeps it on, Rasmr continue to bad mouth him while pumpcade goes to 200m and he just looks sidelined.

[0xProtoCon @0xProtoCon]

@nut_history "the same way I know you're going to meet me for drinks at X at 6pm for happy hour."

Bro blew it.

[0xProtoCon @0xProtoCon]

@HazelAppleyard Assuming you won't die from hypothermia or lack of oxygen, I'll choose 5:30am. Don't need alarm clocks or coffee for 5 years, and $100m should cover the therapy for the PTSD.

[0xProtoCon @0xProtoCon]

@Hoopss Convert it to USDC and put it on Coinbase or similar. Make $2,300 a week in interest. Thanks for playing.

[0xProtoCon @0xProtoCon]

@RaldaMoore @sciencegirl Brother the Windows 95 entertainment pack was absolute fire.

[0xProtoCon @0xProtoCon]

@AirAssets 9 year career. Out now.

[0xProtoCon @0xProtoCon]

Amazon had 20 years to make engineers personally liable for outages. I'm not convinced making them liable just because of AI is even feasible.

You pay an L6 engineer $280k a year plus RSUs to approve changes. How do you make them compensate for a DSI OPS impact of that amount for a single mistake?

Culturally, they'd need to introduce a new Leadership Principle. Are Right a Lot, Earns Trust, and Ownership don't cover "you made an error, caused a COE, and now you're broke."

[0xProtoCon @0xProtoCon]

"And the Ethereum L1 — the thing that gives all of it value — is about to get a lot more powerful." 🤝

[Etherealize @Etherealize_io]

https://t.co/DrKAELg2kw

[0xProtoCon @0xProtoCon]

@TomerAshur In 3-6 month when they have better internal tooling - they'll remove this and no one will talk about it. Temporary at best.

[0xProtoCon @0xProtoCon]

I wouldn’t immediately bet against Amazon’s long term posture. I firmly believe they knew sev1s would happen. COEs, better internal tooling, and giving kira more training data over time directly tied to customer impact and operational metrics are going to make it something not even Claude will have 2-3 years from now - knowing downstream impact on ops and revenue of the generated code.

[0xProtoCon @0xProtoCon]

Amazon has a massive code base. And a ton of engineers. Requiring Kiro use is strategic, if they can convert that usage into a better model. Why pay anthropic margin, who runs on AWS, if they can run Kiro on AWS themselves and charge internal businesses an IMR rate?

The COEs will add process and guardrails to this new regime. I trust the COE process as Amazon does it. It works.

They’ll add bureaucracy. Small changes will be done by humans to expedite. Things will stabilize.

The long term question is Kiro going to be any good mid to long term? Amazon ML has historically worked when predicting your next purchase. It’s been relatively underwhelming on non-objective outcomes.

[0xProtoCon @0xProtoCon]

I worked at Amazon for 9 years. It was amazing to see what tiny, presumably innocuous changes could have massive downstream operational and revenue impact.

AI simply reduces the cognition of its user (since the system designer has been laid off), and increases the number of commits and deploys.

Even when running some of the best teams in my career, more deploys meant more SEV2s. Always.

So increase velocity of pushes with less senior engineers? Ops load.

[0xProtoCon @0xProtoCon]

@sumiturkude007 Everyone nit-picking the imperfections aren't considering where AI video generation was two years ago. Learn to draw lines folks.

[0xProtoCon @0xProtoCon]

@baobanger But there is no war in Ba Sing Se

[0xProtoCon @0xProtoCon]

There's a corollary to this. LLMs haven't given devs better ideas, just lets them ship their bad ideas faster.

As a software consumer, I haven't consumed new services as fast as they've come out, so the bottleneck is very quickly going to be paying customers.

The race to zero on software price is already being seen. Don't think you can make ANY money building a one-click claw deployer, there's already 200 options.

[0xProtoCon @0xProtoCon]

@BitBoyJay Ship it brutha

[0xProtoCon @0xProtoCon]

I feel like I still do a lot of this work but the workflow is different. Instead of talking to myself or thinking quietly I’ll discuss trade offs in realtime with opus, making decisions along the way. Then let it do a first pass, ask a bunch of questions about implementation details, spot and remove duplication or dead code, an then continue to refactor.

The art and craft is still there at least for me, I’ve just eliminated the time spent manually typing.

The code comes out crisp and artful, just as it always has, with great coverage. It’s just faster.

Especially still required on backend systems. Once front ends moved from SSR rails projects to fully client side JavaScript, I accepted that most frontend was disposable slop as long as it worked, knowing it would be iterated on and rewritten anyway.

AI hasn’t changed this, but sped up iteration time.

But it hasn’t made me start to produce slop at areas in the stack it was never welcome.

[0xProtoCon @0xProtoCon]

@SecWar Imagine not reading the room.