So even though claude is better now you still need codex just to read it's output. I'm not reading all that.

New meta: run claude code and codex side by side on the same project and tell codex what claude is doing in the other terminal and ask it "how fucked am I?" so you are emotionally prepared when it breaks everything.

My 2 sats of quantum slop. In Bitcoin, signature schemes are there to authenticate the owner of coins. If there is a quantum computer they don't work. Code that doesn't work should be disabled and removed. Removing broken things is not confiscation but leaving them in is since it allows confiscation by the quantum computer operator. The plan: - Soft fork in SHRINCS into a tapleaf. Require both internal key BIP340 signature + SHRINCS signature. - People slowly migrate to addresses that include taproot + SHRINCS (but don't actually use it to spend other than to test). - Set up 200 bit ECC puzzle as part of bitcoin consensus. If solved, ECC is disabled in Bitcoin and it auto triggers the availability of pure PQ address SHRINCS (or whatever this can be changed as we go). - Encourage the development of prediction markets on the 200bit ECC puzzle so that QC developers have a nice incentive. A liquid or other sidechain prediciton market would be good but you could add an OP on main chain to inspect whether it had been solved or not. - You could add challenges in a range from 150-200 bit to keep things interesting for prediction markets. Solving 160 bit would be a major red flag. Only 200 bit solution triggers the ECC disabling. - After ECC disables, over time plan a hard fork so people who hadn't migrated to taproot+SHRINCS can get their coins back via guy fawkes signatures or ZKPs etc. - sunset the disabling after 40 years (the quantum question is assumed to have been sorted out by then). Why: - QC developers will almost certainly not use their capability to attack bitcoin as a first step. - If honest researchers pull it off we have an incentive which gives them money and prestige - QC may never happen. Sunsets or anything relying on a social assessment is misguided. - In the unfathomably unlikely event that a QC is kept secret and developed just to attack bitcoin AND no honest party can hit 200bit before them then we'll have to manually do the emergency ECC disable softfork and perhaps roll back the chain a bit. Still in a better position since all the code is already in there and most people have migrated to taproot + SHRINCS - Hashed keys are just as vulnerable. The first QC that can solve secp256k1 DLOG given our current state of knowledge will be able to do it quickly -- few seem to understand this. If they are going to attack bitcoin with it they might as well set up a few of them and steal every mempool coin they can to cause maximum panic. Would be better than stealing satoshi's coins.

In the middle of a long and complex task codex decodes to figure out the answer to 1+1. When asked why: "I accidentally invoked the web tool with a trivial calculator call while thinking through the review. It had nothing to do with your repo or the task, and it should not have happened." please un-nerf claude code

And just like that your bitcoin is no longer stored in your house, there's no descriptor backup you can lose and anyone with 2-of-3 devices can restore the funds. Once you obtain this level of peace of mind you can never go back. View quoted note →

npub1r89krrt3u2cugr6aje5n8e0f9jtp6awqj7czhx0t5ga5v5x6gq6s90z95r

52 years later, only known copy of Unix v4 recovered from randomly found tape, now up and running on a system — first OS version with kernel and core utilities written in C

Tom

52 years later, only known copy of Unix v4 recovered from randomly found tape, now up and running on a system — first OS version with kernel and core utilities written in C

A heart-warming story for cold, stony sysadmin hearts.

To all those people who ask me how to get involved in bitcoin open source, here’s the best way:

₿OSS Challenge

₿OSS Challenge - start your career in bitcoin open source

Has anyone tried developing a soft fork that would actually verifiably stop all spam? 1. Make a new address type that has a public key + proof of knowledge. 2.5x larger than current addresses. No script nothing else. 2. disable spending to anything other than the new address type. Spam solved. You could modify (1) to allow raw multisig also. You can actually do lightning without script using MuSig + adaptor signatures. So everything keeps working… in theory :)

I see a few people calling people “unethical” or “dishonest” for their technical opinions. Don’t do this, as good as it feels. Pick up your cross and just make arguments.

“I’m waiting for my bags to have one last pump so I can rotate to BTC” is the best genre of post.

Oh my sides. It's unbelievable that if not for Bitcoin the entire human species would have been forever enslaved to this preposterous system.

Get into it freaks! View quoted note →