Just confirmed that @Ledger is REMOVING the EIP-7702 implementation from its Ethereum app. Everyone deep in Ethereum knows how bad this is 🤦♂️
My PR adding 5 EF-recognized delegators to the app's whitelist waited 9 months for a reply, then got closed.
For Ledger users on @ambire, this means no batching, no gas sponsorship and no paying gas in stablecoins - things the same people have had with a hot wallet for 18 months now, since Pectra.
And when the hardware wallet with the biggest market share drops 7702, the signal is "7702 is unsafe". Its reputation already suffers from misunderstanding - the "7702 drainers" all started with leaked private keys, not with 7702 itself.
@ethereumfndn, this is the last thing 7702 needs right now. Can we channel some energy into getting @Ledger to reconsider? 💪
https://t.co/wfZx0xbCE4
I started learning PCB routing in June
For a month, I was in aggressive denial: “there’s no way a good coding agent can’t do it well”, going down a rabbit hole, developing my own autorouter(s), gradually learning to do it by hand until I became pretty much an expert, and realized progressively how hard this is for an LLM that cannot think in 3D
During that time I still knew AI will most likely solve this problem, cause it’s amazing marketing for anthropic and OpenAI to say they’ve solved it
Now that fable 5.2 has allegedly solved it I can only be happy and curious, I have zero sunken cost regrets. Not to mention being able to verify AI work is extremely valuable for hardware
If your wallet of choice isn’t @ambire and it runs Ledger’s latest signing kit, chances are every signature you make gets reported back to Ledger.
I noticed it early while bumping Ledger packages to debug an issue! No affected version ever reached Аmbire production, and we’re shipping it disabled by default, with an opt-in toggle in the privacy settings 🛡️
However, the devil was that it was sneakily enabled by default in the Ledger SDK since their ledgerhq/context-module v1.16... So right after the lib update - every transaction you sign with a Ledger device started sending to https://t.co/wtXKaYbqGf the network, the contract or recipient address, device model and firmware, app versions and whether you signed blind.
I'd argue that privacy should be the default, @Ledger 💪
Happy to see more and more people becoming aware of the privacy implications and picking Ambire for our ongoing privacy focus.
Catching things like this 👀 is part of the job for us!
Every piece of software you hate was made by humans.
Your bank’s app that logs you out every 90 seconds.
The airline website that crashes at checkout.
The tax portal that only works in one browser.
The printer driver from hell.
The hospital system that loses your records between two buildings.
100% human-made. Handcrafted. Artisanal slop. Written, reviewed, tested and shipped by professionals with years of experience.
And these are the same people now warning us about the quality of AI code.
Brother, we’ve been living inside your code for 30 years.
We know exactly what human quality looks like.
Human slop is infinitely worse than AI slop.
Claude Opus 5.5 can write better code than 99% of developers, yet people who have spent their careers shipping buggy, insecure garbage want to lecture us about code quality.
Maybe you’re not worried about AI slop.
Maybe you’re worried that the “slop generator” is magnitudes better at your job than you are.
there could be exceptions but I feel like this is a red flag, after all software *engineering* is about pushing boundaries with whatever tool is best for the job - and ideally this should be the interesting part
if you're an architect and denounce cranes, you'll be restricted in what you can build (in a reasonable time)
same if you denounce high level languages, same if you denounce AI
software has always been about automating stuff using computers, and now that we're seeing a revolution, ironically so many programmers became disillusioned with coding and jaded ("software isn't interesting anymore because its easy to create")
mf the first computer scientists have DREAMED about a future of programming by natural language. get your shit together.
trying to understand why you say it asymmetrically favors attackers
> But 7702's design is that any delegation can be overridden at any time with a new one.
Ok, but say I get the arbitrary ability to sign - even without 7702 i can flood the mempool with different nonce drain txns. You're gonna have even a harder time stopping me. Esp if i go through flashbots
how was a delegation produced with MPC? Apparently the attacker got a way to sign arbitrary data (akin to the old eth_sign, perhaps the backend was exposing this very method). But logic dictates that if the attacker could sign a delegation, binance could just as easily sign a revocation. The real answer is prob simply that a lot of people are bamboozled due to poor 7702 awareness which bought the attacker time
been increasingly impressed with Ambire
the wallet market is bifurcating imo:
on one side, neobanks abstracting many crypto complexities away
on the other, general purpose wallets built for hardcore onchain users who actually care about self-custody, privacy and control.
Ambire is firmly targeting the latter.
no user tracking. deep commitment to open source. safe multisig by default. and now, a native hardware signer.
what I’m most excited about, though, is their upcoming integration of private transfers.
pre-ordered one
i see people saying
they're moving to iac because agents
they're not using iac because agents
they're switching to a monorepos because agents
they're moving off monorepos because agents
you're all confused man everyone is doing random shit
software is not dead but commoditized
and yes, over time, when building for PmF, a skilled person will do WAY better at steering an AI, precisely because they can assess tradeoffs and know what's easy and what's hard; those things compound
sure, anyone can spawn a Evernote clone, but not everyone can build great stuff
Software is not dead. Here's my unfiltered, honest opinion about using AI for SWE, even in highly complex domains.
It's true that AI is good. Like really good. You can open a complex project in C++, give your Claude Code agent a JIRA task and ask it to create an implementation plan. And it will "almost" work.
As in 95% of it will be correct. So you can literally be a dumb sloperator who just executes commands. But in the remaining 5% it truly matters.
AI can educate you about trade-offs. It can't decide for you. You have to learn that decision-making skill.