malone

paging all devs of farcaster... has anybody been able to get to the bottom of this error?

software engineers, what’s your dev setup going into 2026? I am hearing lots of hyperbole. yes, I know you can run agents while you sleep. but every time you let the ai lead, you lose control which is only regained by thorough verification. I build incrementally with ai, verifying output in lock-step to keep the software tight and clean. times are changing, there is no denying it. over the last few months I’ve been able to continuously loosen the reigns and let the ai cook. I suspect this is going to continue (where it stops I do not know). In the meantime, every engineer is hoping that they’re extracting as much as they can from the models we have access to. Share your thoughts below 👇

Happy new year. Wishing you all a healthy and happy 2026!

Day 54: Batch and bound, we meet again.

Day 53: great listen: https://reason.com/podcast/2025/12/19/hes-serving-5-years-in-prison-for-bitcoin-privacy-software/ we’ve had privacy by default since genesis. Only recently has the word been dirtied by those arguing for more government intervention.

Day 49: reimbursing gas sponsors in erc20 tokens is under utilized in crypto apps today. normalize having not needing eth in your wallet. encourage more apps/wallets to use 7702.

Day 48: r,s,v if you haven't been burned by the 'v' value at least once in your crypto career you're not signing enough txs

Day 47: I once had a colleague early in my career tell me that no problem in our field is unsolvable. The truly hard problems live in research. He also said he’d never faced a problem he couldn’t solve, because when in doubt, he could always reduce it to a state machine. That stuck with me. Fearless debugging is what I want in the people I work with.

Day 46: when in doubt, crack open ethereum magicians https://eips.ethereum.org/EIPS/eip-7821

Day 45: Location change. Back in Ireland for the holidays. @cloaked is taking shape, excited to keep building and share an early version as soon as we can!

the answer is crypto. not the emergent meaning of the word, the original one. the “crypto” that embodies cypherpunk values isn’t expressed through thought pieces or tweeting. it’s expressed through cryptography research and open-source code. the most obvious counterculture movement of the last 10–15 years has been bitcoin and ethereum. if that isn’t obvious yet, it will be soon.

Day 44: Out of all the code I’ve been writing recently, whenever I go back to Solidity and cutting-edge crypto libraries, I find that LLMs have a hard time. In many cases, they aren’t able to arrive at a meaningful solution and I’ve got to do a lot more hand-holding. e.g. Foundry libs, Porto relayers, smart contract assembly.

Day 43: Full transaction history per account is done. It’s been great seeing how fast this came together. I built a similar feature in 2019 and it took about a month. This took roughly a week. Next up is the infra that lets users actually spend. We’ve decided to lean into 7702, so we’ve been surveying who’s making real progress here. Porto keeps standing out. The 'Orchestrator' architecture is nice because it has gas sponsorship, multichain, and multicall baked in. (attached is an out of context sketch from today. I keep doing this to convey ideas and riff on new approaches. hard to beat pen and paper.)

Day 42: today we hit the 3,000 GitHub Actions minutes per month limit. TIL: - parallelism doesn’t save you minutes. it all still counts. - rounding happens per job, not per workflow. so a job that runs for 6 seconds counts as 1 full minute. (this sucks, i don’t make the rules) - you can add an actions budget so that if you spill over 3,000 minutes you start eating into that. - if you want more minutes without relying on budget overages, you need the enterprise tier (50,000 mins/month), but it’s ~5× the cost per user. strategy for now: use the budget and wait for the new month’s allowance to kick in.

Day 41: qr code generation is scary in crypto. i was doing some poking around and it seems like most apps are using https://github.com/kozakdenys/qr-code-styling. if there ever was a critical library to double and triple check, it’s this one. a single nefarious commit and your users are sending funds to an attacker.

Day 40: I bought two domains from EuroDNS two weeks ago. Payment cleared and the money left my account. A few days later they re-listed the same domains for 10× the price I paid. Customer support was useless. After a lot of back and forth, they marked the case as resolved. I re-bought the same domains on Nnamecheap for ~8× less than EuroDNS. Just use namecheap.

https://farcaster.xyz/barmstrong/0x0476d62b coincidence!

Day 39: This is from a tiny sample size and it’s hard to draw any conclusions from the data, but I believe passkeys allow for the best crypto UX today. However, not many people understand the security model they bring. I’m hoping to see more wallets default to passkeys. I get that seed phrases are a necessary evil these days, but I’m really hoping they’re on their way out and only reserved for crypto power users. I like what Porto has been doing in this space. I think it’s viable for a new wallet to solely rely on Porto for account creation.

Day 38: Seeing all this commentary on the Farcaster pivot reminds me of an article Paul Graham wrote in 2009 (https://www.paulgraham.com/identity.html). Most of the takes I’ve read are non-falsifiable. The opinions are based on little or no data and mostly just strongly held beliefs. You’d swear some of these people have built a decentralized social network and are now bestowing their hard-fought knowledge on the rest of us. The reality is they haven’t, and this is a debate where anyone with a strongly held belief can chime in and act like an expert when their opinion can’t be proven wrong by data. These same people are usually the worst offenders of the bike-shedding you see in so many companies today. I trust the @farcaster team to be making the correct, risk-adjusted decisions based on the data available to them. Remember there are many armchair commentators who have nothing at stake when their opinion is wrong.

Check out this poll: "how much of your crypto portfolio would you store on a passkey? "