in a lot of ways, Reylo obviates the entire "ORMs vs SQL" debate ✅ zero codegen ✅ fully inferred: totally DRY ✅ maximum expressiveness: CTEs, window fns, subqueries...the full power of SQL and with agents, it doesn't matter that SQL syntax is a monstrosity 😅

AI agents are writing more code. How can developers ensure that code is correct? Microsoft Research’s Proof-Driven Development (PDD) Kit explores a more rigorous approach: it translates a developer’s intent into a formal specification, then uses formal verification to prove that generated code meets it. PDD Kit works with Verus for Rust and is integrated with @GitHubNext’s Chopin, bringing formal methods into an AI-assisted development workflow for teams. See the work in action at @github Universe, Oct. 28–29, 2026.

in the next version of Porffor, Test262 compliance leaps to 96%! this marks a milestone where Node compat and performance finally become major blockers

I'm an agentic engineer. I strive to gain confidence in my ability to judge verification evidence from my agents to push forward ambitious complicated projects to completion with minimal human involvement. In the past week Opus 5.5 pushed 7 tax returns/amended returns to completion for me, across multiple people, years, and states. I signed each return or handed over each return with very high confidence in correctness. Opus also found errors from previous tax professionals. I historically have been an extremely meticulous reviewer of any tax return that I must sign. I believe in being honest and acting with integrity. I often find various errors in every return that any tax professional has ever prepared for me. Usually 5-10 errors, most immaterial but wrong, but usually 1-3 material errors. Sometimes I have caught horrific errors that would have been completely illegal! You need to know how seriously I take any tax preparation I'm involved in. In the past any given year or returns probably took me 20-40 hours of personal work on top of any professional helping me. Also, working with these professionals has generally been a terrible experience. And I have worked with many over the years both professionally and personally, spending probably around $40-$50k in the last 6 years combined. These professionals overlook things, don't do a meticulous review to my standard, which is to do a meticulous line by line review. I always caught things that they would have caught if they were reviewing like me. I'm not even a tax professional. My professionals would often get upset or impatient with me because of my demands of them providing me with a near 100% accurate return, which to me is crazy because I'm paying for this service. Well, all of that is over now. Done. I don't foresee me ever paying a human being to prepare my taxes again. I've been using AI to aide me in tax preparation since probably 2022. At first it was so unreliable it was barely helpful. It made very basic arithmetic errors that just didn't allow you to trust it. That went on probably from 2022 to 2025. Near the end of 2025 things got way way way better. AI began to do very in-depth reviews of professionally prepared returns. Around October 2025 it caught an absolutely fundamental material mistake that our preparer had made...we weren't even legally allowed to amend our returns because of the BBA! We had to do AARs. Our preparer had completely overlooked that and had provided amended returns and we were about to file them. The preparer agreed, I mean it was irrefutable once you understood the law. In December 2025 I used AI to drive forward a complicated AAR...the first time in years that I was attempting a complicated return without the help of a professional. Wow...it took a lot of crazy guidance and generated code and intense reviews from me...but within 24 hours (like literally I worked with the AI for about 24 hours straight) I had a final AAR prepared, meticulously reviewed, and I felt confident enough to sign it. You must understand, if I am signing it, that means it is a return of an extremely high quality. I feel an incredible aversion to signing something that I do not feel confident in, and like I said I always find many errors across professionally prepared returns. This feat in December was an amazing unlock for me. Oh yeah, and our tax preparer basically had already fired me a month or two before, due to his inexperience with the AAR process and probably due to my difficulty as a client in demanding meticulous excellence. So, December was the first proof. I was no longer bound to pay $1000s to navigate this ridiculous system to do my duty to my government. A huge burden began to be lifted from my shoulders. Taxes are probably the most terrible thing in my life that cause me extreme anxiety and hardship. That first AAR was done with GPT-5.2 Codex I believe. Math seemed to no longer be an issue. In March I did it again with an entire partnership return. In the past, professional, $1-$3k probably. GPT-5.3/5.4 Codex I believe. Done. Finished. PDFs entirely filled out even. The best tax preparation experience I had ever had, the AI treating me as I felt a tax professional should always have been treating me. Instant answers, always available, intense levels of research, as much meticulousity as I want...explaining trade-offs, writing down decisions with evidence...just an amazing experience. And then I did it again with a personal return in July. And this past week I did it again with 7 more returns or amended returns. This past week I used Opus 5.5 in the Claude Code desktop app. It was mind-blowing. It was so amazing. I think it only maybe like 5ish mistakes, maybe one material that it caught...the rest were very very nitpicky. Towards the end of the week I was getting so confident that I really stopped the line by line review process for most of the review. I guided Opus 5.5 to engage in multiple targeted intense reviews of its own across schedules and certain lines, and I judged its evidence. I somewhat went line by line through the 1040 for example, but left most of the schedules to a higher level review. Our work, decisions, evidence, everything is kept organized and available to the AI in GitHub repos. Once I sign something I send it over and it gets organized, committed, and pushed automatically. It tracks tracking numbers from certified mail receipts. It tags and releases the final verified return. This is one of the most amazing technological experiences of my life. I will never pay for a tax professional again, is my plan. And I hope many more people will stop paying for tax professionals to do this work. I don't believe that most tax professionals do great work. I'm sorry, but I believe it's true. I've experienced too many of them to know the level of work they produce. And unfortunately it's somewhat inherent to the human condition and the monster that the government has created. Tax preparation is a huge drain on society, producing work of very little value to human progress and flourishing, just to calculate a couple numbers to tell you how much money you can't keep. This is an outrageous drain on society! These people should be engaged in actually productive work pushing society forward. The government, IMO, has stolen 100s of hours of my personal time, $10ks of dollars paid out to just prepare these things, and my internal peace. This must end, and AI will end it. I encourage everyone to begin to embrace the super intelligences that are coming. Use them to cut out human suffering and wasted time and money.

Landing improvements now, it’s actually quite tricky to get this right and we’ve been doing a fair bit of science here. Will share more soon.

Today I call upon the blockchain industry to calmly begin planning for "bunker mode". My personal recommendation is to set in motion a controlled mass migration of assets to fresh addresses, i.e. addresses whose pubkeys remain hidden behind a hash. Holders, starting with large and sophisticated ones, should consider moving the bulk of their funds to addresses that have never signed a transaction. And when they do sign one, they should also move remaining funds to a new address (possibly generated from the same seed phrase). Don't rush. While I believe there is cause for action a rushed migration would do more harm than good. Don't panic either. Moving assets to protected addresses is a simple, preventative step which does not require new cryptography or new wallets. IMO it is now reasonable to brace for the possibility that ECDSA breaks before qday, in the worst case in months not years. By "break" I mean fast private key recovery (e.g. in one week) on available hardware (e.g. a large GPU cluster). Recent days have been humbling for human mathematical intuition. Long-held, unquestioned hypotheses have fallen. This includes the n log(n) bound for integer multiplication and the 3SUM conjecture. In hindsight, May's unexpected disproof of the Erdős unit distance conjecture was our warning shot. Yesterday's OpenAI drop made it clear that mathematical superintelligence is upon us. They say there are weeks where decades happen. We are about to live through weeks where centuries of mathematical progress happen. Could our magic 64-byte ECDSA signatures be too good to be true? Was it just security through obscurity all this time? Elliptic curves feel especially vulnerable to superintelligence. Curves carry rich structure, with room for fancy tricks like Schoof, Frobenius, pairings. (By contrast, hashes are designed to minimise algebraic structure.) Separately, as Ewin Tang can attest, an efficient quantum algorithm sometimes foreshadows an efficient classical one. We should be open to the possibility of a classical counterpart to Shor that breaks elliptic curves and RSA at once. Also noteworthy is the striking under-representation of cryptographic breakthroughs among the 722 mathematical results OpenAI published. I've witnessed first-hand the US government censoring academic quantum cryptanalysis results. Backroom interventionism is my base case. I urge large, sophisticated actors to lead by example. Project11's "risq list" (bitcoin-risq-list.projecteleven[.]com) is a great tracker of exposed BTC pubkeys. Binance, Bitbank, Robinhood, Bitfinex, and Tether have an opportunity to harden their cold storage. Next month I'll address institutions in London in a live Q&A (forum.ethereuminstitutional[.]org/london-2026). Again, please do not rush. Wallets holding under 50 BTC enjoy partial cover from "Satoshi's shield", i.e. his 20K exposed addresses that hold 50 BTC each. Load-bearing signers like oracles and L2 security councils should consider rotating ECDSA pubkeys with every signed message and/or multi-signing with a hash-based schemes like SPHINCS. Exiting bunker mode safely will require post-AI cryptography. My inclination is to go all-in on hash-based cryptography and avoid structured mathematical assumptions entirely, whether from curves, lattices, or isogenies. A single battle-tested hash (e.g. from the SHA or BLAKE families) yields plausible post-AI security. The Ethereum roadmap on strawmap[.]org fully embraces hash-based cryptography with end-to-end formal verification as a response to the quantum threat. Those timelines must now be revisited and accelerated in light of mathematical superintelligence. I'll be pushing for maximum defensive acceleration.

Hot take: You can vibe-code a small app. You can't vibe-maintain a 90K+ line codebase. No chance.

I asked Opus 5.5 to write, produce, and create me an educational 'cartoon' about The History of Light - using whatever resources it had available to it. I didn't expect it to make a 36 minute video and spend $200+ dollars though 😂 The study of light is genuinely fascinating, and we have spent thousands of years trying to understand it. This is that story. Opus 5.5 wrote and produced the entire film. I told it to use any keys in my keychain to generate the voices, create any images or video it might need, and to create whatever motion graphics it could to explain concepts in electromagnetism. It used > Gemini 3.8 Flash TTS for the narration, and I must say, it did an excellent job. > Lyrica 3 for the background music, which is also essentially perfect. > GPT Image 2.5, Nano Banana Pro and NB2 Flash for some of the scenes imagery It's certainly not perfect but I was still really impressed with how well it did producing the film. It's very interesting if you're a physics nerd like me. Opus is genuinely creative in a way that I've not really seen with other models. I just gave the model a few constraints and just let it make all the decisions end to end. It even wrote itself credits at the end lol. It's going to be so fun to create bespoke educational videos like this for my children to learn from

not gonna lie, my codex dot is pretty useless so far - it fails at almost everything. even creating a simple apple reminder didn't work: it went silent for hours, then randomly told me an error happened - i asked it to check showtimes for a movie night, it took ages to find them, then an hour later messaged me to say it got several things wrong - calling it works better, though it takes 10-15 seconds to pick up, and then it tends to go silent until i re-ask something after patiently waiting for 10+ minutes - the only things that worked well were checking my gmail inboxes and setting recurring reminders - i'm not sure what the cloud computer is for, any ideas? is it just a persistent chat thread with a cute blob and a hidden memory.md it updates as you talk? someone please help me understand, i'm utterly confused...