Paweł Bylica

AI meets ZK: "Make comments succinct".

@OisinKyne @nero_eth @M25Marek Not this year.

@nero_eth @M25Marek How do we get a prediction market for when glamsterdam ships on mainnet live? :)

It's been 210 days since Fusaka activated, and we're getting closer to the next fork: Glamsterdam. Since the Merge, Ethereum has shipped a fork every ~294 days on average. Shapella: 210 days Dencun: 336 days Pectra (pls don't remind me): 420 days Fusaka: 210 days Glamsterdam is shaping up to be one of the bigger forks. ePBS revamps the block validation pipeline while reducing trust assumptions. BALs bring parallelization to Ethereum, and repricing EIPs make sure we can keep scaling safely. On top of the headline features, we're also getting ETH transfers emitting logs, a SLOTNUM opcode, a deterministic deployment factory, larger contract code size, and more. Glamsterdam will scale Ethereum by around 5x, and with Hegota and a few more repricing EIPs, another 2–3x is within reach. At the same time, work on Hegota is already ramping up. Inclusion Lists (FOCIL) + Frame Transactions will be a great combo, and there are already 20+ EIPs proposed alongside the two headline features. Further out, zkEVMs are the next big theme. PQ Ethereum is also on the horizon, and there's already work underway to get the protocol ready for it. Before we get there, there are still things to do: revamp state and sync, improve RPCs, ship faster finality, shorten slots, add native privacy, and more. Lots of stuff ahead, but it's good to see Ethereum's roadmap starting to come together.

@leonardoalt I want to learn FV for EVM. What does it mean to formally verify an EVM implementation?

I'm back to FVmaxxing. DMs and comments are open if you want to discuss/learn/shill FV for smart contracts/compilers/ZK, Lean, FV+AI (!!!), SMT, evm-smith, or anything related. I also updated github.com/leonardoalt/et… with newer info, happy to take more contributions!

@pcaversaccio @Mudit__Gupta @omw_to_the_moon EVM stack exists.

well i think this analogy is flawed: MCOPY & CLZ optimise ops that were already semantically correct but expensive on-chain. stack-too-deep is a _compiler artifact_: what i mean by that is it does _not_ exist at the EVM level at all. spilling to memory solves it w/o touching the base layer at all.

Glmasterdam is lining up to my favourite hard fork till date 3 EIPs that will completely change the devex of L1 EVM. - Increase code size limit - 2 new opcodes that solve stack too deep - ETH transfers emit log Alt L1 chains and EVM L2s - for your sake - I hope you take a look!

@pcaversaccio @Mudit__Gupta @omw_to_the_moon Sure, spilling to memory works, but it’s a bad trade-off. By that logic, MCOPY is unnecessary because bytes can be copied one at a time, and CLZ is unnecessary because it can be emulated with ~60 existing EVM instructions.

@Mudit__Gupta @omw_to_the_moon no it's not. any new addition brings a burden for execution clients and QoL improvements have no place there. new opcodes/precompiles _must_ be the absolute last solution to any problems and this "problem" has been already solved long time ago for compilers via stack spilling.

@barnabemonnot @DefiIgnas LFG!

@DefiIgnas fastconfirm.it "And it's dead annoying though that full ERC-20 deposits to CEXs still take ~13 minutes (no 1-slot confirmation)" Small nugget in your larger post but this is something that is at hand, see this fast confirmation +1 on your larger point around UX

$ETH went from a consensus hold to a contrarian bet in 2-3 years. Some of this was market driven, some was self-inflicted: 1) The EF pushed the L2 scaling roadmap after failing to scale the L1. It doesn't matter whether that came from lack of motivation, skills, or available tech and research at the time. Now the EF is scaling the L1, but even with gas limit increases, Ethereum will never be faster or cheaper than most competitors. And that's okay, because maximizing decentralization and censorship resistance requires tradeoffs. The problem is that market participants are giving it lower valuation multiple than in the past. And it's dead annoying though that full ERC-20 deposits to CEXs still take ~13 minutes (no 1-slot confirmation) and that approve + action still requires two txs across DeFi, despite years of 'account abstraction' upgrades. Watching EF members leave one by one isn't helping the sentiment either. 2) Ethereum can be slower and pricier than other chains, but the market now wants revenue to back valuations. $HYPE is generating 2x-3x the fees of Ethereum despite trading at ~5% of its market cap. Even more humiliating is $TRX, up 5x while ETH is down 40% over 5 years. Ethereans mocked TRX as a copy/paste vaporware scam, but Tron dominates retail stablecoin payments... The sector EF pushed for years and failed to capture, because Ethereum was simply too expensive and slow for adoption. Ouch. I believe Ethereum had it good with the ultrasound money narrative. Quickly deflating supply is the sexiest narrative that even BTC bulls would love. But it needs a massive pick up in txs numbers to generate the fees that burn ETH. And Glamsterdam just cut fees by ~78% (gas limit will go from 60M to 200M per block), which means transactions need to pump by 4.6x just to keep the burn flat. If onchain activity doesn't pick up to compensate, Ethereum's revenue drops further. Sure, Ethereum still dominates TVL but the ratio dropped from 96% in Jan 2021 to 52% today. And even with that, TVL monetization mostly flows to protocols and stablecoin issuers, not the L1. L2s aren't taxes either. ---- So what's the bullish case for Ethereum here? EF has partly got the message. The cypherpunk manifesto is personally very appealing to me, with its mission to promote privacy, self-sovereignty, and independence in an increasingly unstable world. I hope that recent departures from the EF is simply a realignment period. Pivoting to L1 scaling is the also right move, but UX needs to drastically improve, especially as more corpo-slop chains and institutions enter the market. EF is taking the quantum threat seriously, unlike the mixed reaction from Bitcoin core devs. But that all takes time, and if the market's demand for revenue doesn't subside, Ethereum simply needs to bring more users and transactions to the chain. The real ultrasound money narrative, while being the most decentralized chain, would do the trick. But we're far from ETH being deflationary again.

@zeddotdev @danilobleal Too slow to be usable on bigger projects (e.g. LLVM).

Oops, we merged like, 7 new Git features in the last few weeks! ¯\(ツ)/¯ Shoutout to @danilobleal for the new History tab in the Git panel, landing in stable this week.

@ben_a_adams @InstLatX64 Like EVM.

@InstLatX64 Keccak1600 does want a cursed number of registers so likes to spill to stack (also problematic for SIMD). Not sure if by design or accident but APX should help a lot there (and for other stack spilling algos)

This is an important question, and now I finally have some static data: #Intel #APX has the shortest code path, at least in this hand-optimized #SHA3 / #Keccak1600 case

@gakonst Just use LLVM's musttail marker. #call-instruction" target="_blank" rel="nofollow noopener">llvm.org/docs/LangRef.h…

how can we stabilize tail call optimization in rust ASAP? we're happy to help fund it if necessary. can anyone connect me to the right people please?

@gakonst This is the Global Seed Vault on the picture. It has no code.

the reth team just returned from interop preparing for the Glamsterdam hardfork L1 scaling looking bright with ePBS, BALs, gas repricing and snap/v2 they also got to see the arctic code vault!

Last week we trespassed into the realm of RTP with 2 GPUs, this week we broke the 8 second barrier with zilkworm-airbender @eth_proofs Next-stop: Avg latency less than 10 seconds

@leonardoalt Some time ago AI had successfully corrupted my 2 independent git projects.

AGI

Fil-C is a C/C++ compiler that builds C/C++ projects into a binary with "complete memory safety". It is a great project by @filpizlo. In part because it is so easy to do, I am going to make sure that all the production-quality projects I manage support Fil-C directly. For the most part, this means skipping over direct assembly calls. You can still have the assembly calls, but you need some alternatives code path when building with Fil-C. As a side benefit, running the code with Fil-C serves as a sanitizer: if there is any bad memory business going on, Fil-C will just crash you out. Interestingly, Fil-C has no escape hatch. It is a totalitarian approach.

Researchers just proved that every single elementary function, sin, exp, log, sqrt, comes from one single binary operator. It is like finding the “God Particle" for calculus. In computer science, every complex program breaks down to a single logical operator: the NAND gate. It is the fundamental building block of all digital reality. But for continuous math, physics, engineering, machine learning, we thought we needed a massive toolbox. Addition. Subtraction. Trigonometry. Logarithms. Every scientific calculator and neural network has to juggle all of them. Until today. But this paper proved that every single mathematical function can be generated by a single, bizarre binary operator. eml(x,y) = exp(x) - ln(y). Combine that with the number 1, and you can build everything. Pi. The square root. Sine and Cosine. Arithmetic. It is all just the exact same operator, repeating over and over again in a binary tree. Nobody anticipated this existed. It was found by systematic exhaustive search. But the implications for AI are massive. Instead of an AI struggling to combine different mathematical rules to discover a new scientific law, it can just use a single, uniform architecture. One trainable circuit. One repeatable node. We thought the language of the universe was complex. It turns out, it's just one equation repeating in the dark.

GPU prices aren't gonna like this. Sub-12 second #etheruem #zkevm proving achieved on just 2 5090s ethproofs.org/clusters/5a74e… As a great man once said, "basically a toaster"

In the future, EVM will spend more time computing gas than computing.

Announcing powdr-wasm! powdr-wasm is an optimized zkVM for WASM, built on top of @openvm_org and the novel 𝑐𝑟𝑢𝑠ℎ ISA. Early benchmarks already show 1.5x fewer trace cells & faster proof times compared to RISC-V (OpenVM). It also supports Go guests via WASI! 👇

@adrianbrink We have CLZ now.

I get the overall vibe, but Ethereum isn't even friendly to developers as a protocol. We had an EIP-665 (eips.ethereum.org/EIPS/eip-665), which was implemented in every client to add ed25519 precompiles. However instead of doing the simple thing, someone proposed to generalize it (ethereum-magicians.org/t/precompile-f…); make it work for all curves. So as a result the former was abandoned and closed and then the latter was also abandoned and closed. The net result is that 8 years later, we still only have secpk1 and bn128 precompiles instead of at least also having ed25519, which everyone agrees is just better. So I get why people complain that this looks not focused enough on actual users, sure. But god the problem is that this is also not focused on any protocol user. There are a bunch of super low hanging fruits (like freaking precompiles) that would make deploying privacy tech 10x easier, but we haven't managed to improve the core protocol a single bit on that front for the last 8 years.

@big_tech_sux You just want the "stack too deep" to stay so you can point it out in every Solidity vs Vyper discussion.

eip-8024 is a pretty useless eip, there are much better and more useful improvements for the evm like fixing the 70s memory model, why are the core teams prioritizing this!!