ACDE2026-10-08_247

AllCoreDevs - Execution #247

2026-10-08 5 decisions 658 transcript lines

Transcript

Follow along

Kevaundray Wedderburn
gm
Justin Traglia
gm
read.ai meeting notes
Ignacio added read.ai meeting notes to the meeting. Read provides AI generated meeting summaries to make meetings more effective and efficient. View our Privacy Policy at https://www.read.ai/pp Type "read stop" to disable, or "opt out" to delete meeting data.
Benedikt Wagner
gm
ethDreamer (Mark)
gm
jochem-brouwer
gm
potuz
gm
spencer
gm(cubed)
marc | wolovim
Replying to "gm(cubed)" gm (youre unmuted)
Justin Traglia
And boring is good 🙂
Kevaundray Wedderburn
Boring == good
spencer
Reacted to "gm (youre unmuted)" with 😁
Mario Vega
Reacted to "And boring is good 🙂" with ❤️
nixo
Reacted to "And boring is good 🙂" with ❤️
Lea Na
A réagi à "Boring == good" avec ❤️
spencer
Reacted to "And boring is good ..." with ❤️
Justin Traglia
Reacted to "Boring == good" with ❤️
nixo
Reacted to "Boring == good" with ❤️
Iván | ethrex
No news good news
potuz
Wasn’t that the update?
Luca Donno | L2BEAT
Reacted to "Boring == good" with ❤️
Pooja Ranjan
Reacted to "Boring == good" with ❤️
Iván | ethrex
Reacted to "Boring == good" with ❤️
Luca Donno | L2BEAT
Reacted to "And boring is good 🙂" with ❤️
Barnabas
Reacted to "And boring is good 🙂" with ❤️
Justin Florentine
ePBS + boring fork == accomplishment
Kevaundray Wedderburn
Reacted to "And boring is good 🙂" with ❤️
spencer
Reacted to "Boring == good" with ❤️
felipe
Reacted to Boring == good with "❤️"
Stefan Starflinger
Reacted to "Boring == good" with ❤️
Justin Florentine
but whats the builder/searcher participation?
Nico Flaig
Reacted to "Boring == good" with ❤️
potuz
Bombora
Daniel Knopik
2 are sigma prime
pk910
likely bids don't pay more than local blocks
Dustin
are the builders actually making better blocks
Nico Flaig
I am just bidding with local payload value
potuz
Yeah that’s the case for the pool I saw
Justin Traglia
Reacted to "I am just bidding with local payload value" with 👍
nixo
https://github.com/ethereum/pm/pull/2263 Mon, 26 Oct 2026, 17:42:48 UTC Tue, 27 Oct 2026, 21:01:12 UTC
Nico Flaig
adding subsidy now so we can see
Barnabas
Replying to "likely bids don't pa..." ah right, nobody is spending extra nor mev
Justin Traglia
Btw here’s the agenda: https://github.com/ethereum/pm/issues/2234
nixo
Reacted to "Btw here’s the agenda: https://github.com/ethereum/pm/issues/2234" with 🙏
Iván | ethrex
Could we do it on Tuesday?
james
Do we need 2 weeks before fork for release?
potuz
When do we need releases out though?
Enrico Del Fante
monday
Miguel Tenorio
Tuesday looks like the lucky day for forks tho
Barnabas
https://github.com/eth-clients/hoodi/pull/31
james
That’s pretty tight 🙁
james
Yeah we have a lot of prs to merge in
Barnabas
we still have 95% controlled validator set
FLCL (Nethermind)
An early shadowfork mb?
Justin Traglia
Could we target Oct 13th for releases?
nixo
Replying to "Could we target Oct ..." i think they're saying that's a lil tight
Parithosh Jayanthi
Reacted to "we still have 95% controlled validator set" with 👍
Dustin
that's a nonstarter
Dustin
tomorrow isn't happen
james
Replying to "Could we target Oct ..." Oct 13 feels really early, we’ll have to basically rush to merge things and miss a lot of fixes for hoodi
james
Replying to "Could we target Oct ..." We need 3 days to release
Dustin
Replying to "Could we target Oc..." Yeah similarly, Nimbus tries to have ~no (interesting) changes for 3 days
Justin Traglia
Replying to "Could we target Oct ..." Hmm okay, so maybe Friday October 16th is fine then. I think 10 days is sufficient.
Nico Flaig
Replying to "likely bids don't pa..." I am building most of sepolia payloads now
Barnabas
Replying to "Could we target Oct ..." I really think 10d is sufficient
felipe
How can we coordinate these release updates better? That's worth thinking on.
Barnabas
Replying to "likely bids don't pa..." are you subsidizing?
Justin Traglia
Reacted to "I really think 10d is sufficient" with 👍
james
I thought we always agreed to give 2 weeks for operators on the docs
Lean F | ethrex
On Tuesday we’ll have less people from the team on a plane
Nico Flaig
Replying to "likely bids don't pa..." yes so I am paying more than local now for sure
cayman
Reacted to "Hmm okay, so maybe F..." with 👍
Justin Traglia
I prefer a slightly shorter heads up (10 days instead of 14) if the alternative is to delay the upgrade.
Nico Flaig
do circuit breakers work?
james
Ours is not merged
Justin Florentine
it's doable for us
potuz
Replying to "do circuit breakers ..." I think Prysm’s P2P works, we haven’t tested really the relay one, need to flip a flag and ship it to devnet 8
Enrico Del Fante
We can do that
Łukasz Rozmej
we can release config only changes at any point
potuz
You guys are rich you can pay flight internet
Barnabas
lot more people would be flying on tuesday tho
CPerezz
Reacted to "You guys are rich yo..." with 😂
Luca Donno | L2BEAT
Reacted to "You guys are rich you can pay flight internet" with 😂
Justin Traglia
Reacted to "You guys are rich you can pay flight internet" with 😂
Stefan Starflinger
Reacted to "You guys are rich you can pay flight internet" with 😂
Łukasz Rozmej
Reacted to "You guys are rich you can pay flight internet" with 😂
james
Reacted to "You guys are rich you can pay flight internet" with 😂
Nico Flaig
Reacted to "You guys are rich yo..." with 😂
felipe
Reacted to You guys are rich yo... with "😂"
Barnabas
Reacted to "You guys are rich you can pay flight internet" with 😂
Akash | ECH
Reacted to "You guys are rich you can pay flight internet" with 😂
spencer
Reacted to "You guys are rich ..." with 😂
Barnabas
Replying to "Ours is not merged" merge it 😄
pk910
Reacted to "You guys are rich yo..." with 😂
Enrico Del Fante
Reacted to "merge it 😄" with 😂
Mario Vega
Reacted to "You guys are rich you can pay flight internet" with 😂
spencer
has anyone actually had a good plane wifi experience, it always sucks for me
marc | wolovim
Put your itinerary and SSN below
james
Reacted to "merge it 😄" with 😂
Enrico Del Fante
We are not flying on mon nor tue
Stefan Starflinger
Replying to "You guys are rich you can pay flight internet" ask pari
cayman
Lodestar team will be meeting in person that week, not flying
Justin Traglia
Replying to "You guys are rich yo..." Starlink 🙂
potuz
These people take a lot of vacations I’m only flying like the 30th
jochem-brouwer
i will come by bike
Łukasz Rozmej
Reacted to "Starlink 🙂" with 💯
spencer
Reacted to "Starlink 🙂" with ❤️
Stefan Starflinger
Reacted to "Starlink 🙂" with 💯
felipe
United has Starlink now apparently
Barnabas
Reacted to "i will come by bike" with 😂
felipe
Reacted to Starlink 🙂 with "💯"
Mario Vega
Reacted to "i will come by bike" with 😂
potuz
Reacted to "i will come by bike" with 😂
Enrico Del Fante
Reacted to "i will come by bike" with 😂
CPerezz
Replying to "These people take a ..." You’re not valued enough
Pooja Ranjan
Reacted to "i will come by bike" with 😂
Iván | ethrex
Reacted to "i will come by bike" with 😂
Stefan Starflinger
Reacted to "i will come by bike" with 😂
Luca Donno | L2BEAT
Reacted to "i will come by bike" with 😂
Nico Flaig
Reacted to "merge it 😄" with 😂
CPerezz
Replying to "These people take a ..." #vacationDAO
spencer
Reacted to "i will come by bike" with 😂
Enrico Del Fante
Prefer mon for the time
FLCL (Nethermind)
Reacted to "#vacationDAO" with 🙏
Daniel Knopik
Reacted to "#vacationDAO" with 🙏
Justin Florentine
28/29 are hotspots for besu/teku
Miguel Tenorio
Reacted to "Prefer mon for the time" with ➕
Louis
Replying to "You guys are rich yo..." ^ Let’s cool, is it stable
Louis
Replying to "You guys are rich yo..." *That’s
potuz
Reacted to "#vacationDAO" with 😃
Parithosh Jayanthi
Reacted to "You guys are rich you can pay flight internet" with 😂
Justin Traglia
Reacted to "Prefer mon for the time" with ➕
Justin Traglia
Reacted to "i will come by bike" with 😂
Parithosh Jayanthi
Replying to "You guys are rich yo..." Im flying on united right now and attending ACD via starlink. The future is here and I didn’t even need to pay for it 😄
Łukasz Rozmej
Reacted to "Im flying on united right now and attending ACD via starlink. The future is here and I didn’t even need to pay for it 😄" with 👍
Barnabas
Reacted to "Im flying on united right now and attending ACD via starlink. The future is here and I didn’t even need to pay for it 😄" with 🥦
Pooja Ranjan
Reacted to "Im flying on united right now and attending ACD via starlink. The future is here and I didn’t even need to pay for it 😄" with 👍
felipe
Replying to "You guys are rich yo..." No idea. I had it on a domestic flight recently and it was free but I have a 4 month old daughter so......
felipe
Reacted to Im flying on united ... with "👍"
Kamil Salakhiev
Reacted to "Im flying on united right now and attending ACD via starlink. The future is here and I didn’t even need to pay for it 😄" with 🥦
spencer
Reacted to "Im flying on unite..." with 🥦
Luca Donno | L2BEAT
Reacted to "Im flying on united right now and attending ACD via starlink. The future is here and I didn’t even need to pay for it 😄" with 🛰️
felipe
Replying to "You guys are rich yo..." No testing was done 😆
spencer
Reacted to "No testing was don..." with 😅
spencer
Reacted to "Im flying on unite..." with 🛰️
felipe
Reacted to Im flying on united ... with "🔥"
danceratopz
Reacted to "Im flying on united ..." with 🔥
spencer
Reacted to "#vacationDAO" with 😃
nixo
Reacted to "IMG_9362.jpeg" with 😄
Ben Edgington
Reacted to "IMG_9362.jpeg" with 😄
Parithosh Jayanthi
Reacted to "#vacationDAO" with 😃
Barnabas
Reacted to "IMG_9362.jpeg" with 😄
Pooja Ranjan
Reacted to "IMG_9362.jpeg" with 😄
felipe
Replying to "You guys are rich yo..." Amazing! Finally feels like we live in 2026 lmao
felipe
Reacted to IMG_9362.jpeg with "😄"
Miguel Tenorio
Reacted to "IMG_9362.jpeg" with 😄
Kamil Salakhiev
Reacted to "IMG_9362.jpeg" with 😄
Daniel Knopik
Reacted to "IMG_9362.jpeg" with ❤️
Miguel Tenorio
Reacted to "Amazing! Finally feels like we live in 2026 lmao" with ➕
Pooja Ranjan
Reacted to "Amazing! Finally feels like we live in 2026 lmao" with ➕
Louis
Thank you Stefan!
Ben Edgington
Reacted to "Im flying on united right now and attending ACD via starlink. The future is here and I didn’t even need to pay for it 😄" with 🔥
Parithosh Jayanthi
Reacted to "Amazing! Finally feels like we live in 2026 lmao" with ➕
Will Corcoran
Replying to "You guys are rich yo..." We were promised 140 chars and we got flying zoom mtgs
soispoke
A réagi à "i will come by bike" avec 😂
Parithosh Jayanthi
Reacted to "We were promised 140 chars and we got flying zoom mtgs" with 🫡
felipe
Reacted to We were promised 140... with "🫡"
cayman
Reacted to "Im flying on united ..." with 🔥
Parithosh Jayanthi
Replying to "You guys are rich yo..." I’m so sure in a few years we’ll have call booths in flights
spencer
Reacted to "IMG_9362.jpeg" with 😄
Jihoon
Reacted to "Im flying on united right now and attending ACD via starlink. The future is here and I didn’t even need to pay for it 😄" with 👍
Luca Donno | L2BEAT
Reacted to "#vacationDAO" with 🙏
Stefan Starflinger
Reacted to "Im flying on united right now and attending ACD via starlink. The future is here and I didn’t even need to pay for it 😄" with 👍
Tuyen
Reacted to "Im flying on united right now and attending ACD via starlink. The future is here and I didn’t even need to pay for it 😄" with 🔥
felipe
Reacted to i will come by bike with "😂"
danceratopz
Retrospective EIP Complexity Assessments - Slides: https://danceratopz.github.io/retrospective-eip-complexity/slides/ - Comparison results: https://danceratopz.github.io/retrospective-eip-complexity/results/predicted-vs-observed/
Stefan Starflinger
I heard that the 26th is a public holiday here in Austria :kek:
potuz
Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead
ethDreamer (Mark)
Reacted to "Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead" with ❤️
Justin Traglia
Reacted to "Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead" with 😅
soispoke
A réagi à "Let’s ask the LLM..." avec 😅
Satyajit
Reacted to "Let’s ask the LLM..." with 😅
Will Corcoran
Reacted to "Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead" with 😅
felipe
Reacted to Let’s ask the LLM if... with "😅"
Greg K | Lido
Reacted to "Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead" with 🌶️
Luca Donno | L2BEAT
Reacted to "Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead" with 🌶️
Csaba
Replying to "Let’s ask the LLM if..." Or just ask it to develop everything missing for Q-day and an PQ ethereum in parallel.
felipe
Reacted to IMG_9362.jpeg with "❤️"
Mario Vega
Reacted to "Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead" with 🌶️
Will Corcoran
Reacted to "Retrospective EIP Complexity Assessments - Slides: https://danceratopz.github.io/retrospective-eip-complexity/slides/ - Comparison results: https://danceratopz.github.io/retrospective-eip-complexity/results/predicted-vs-observed/" with 🫡
james
Reacted to "Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead" with 😅
nixo
Reacted to "Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead" with 😅
nixo
Reacted to "Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead" with 🌶️
NC
Is AI accurate on estimating? Sometimes AI quotes like 2 - 7 days for building something, but it actually took 20mins to finish everything. This is engineering hours btw
marc | wolovim
Replying to "Retrospective EIP Co..." Human evals also visible here: https://forkcast.org/upgrade/hegota/test-complexity/ (some hidden behind the “Active only” filter)
Justin Florentine
Replying to "Let’s ask the LLM if..." please no i'm already too future shocked as it is
Csaba
Replying to "Let’s ask the LLM if..." Just tell it: “Make sure PQ Ethereum is finished first”, and you should be fine
Will Corcoran
Human writes
felipe
Replying to "Let’s ask the LLM if..." If we put all our tokens together we can have one mega fork implemented by Claude in all clients and we all just focus on tests 😜
james
Going to be hard to estimate, it assumes team resources and skillsets are the same 🤔
felipe
Reacted to Just tell it: “Make ... with "✅"
Lea Na
A réagi à "Let’s ask the LLM..." avec 😅
Trent Van Epps
Replying to "Is AI accurate on es..." Shouldn’t the relative measure between features be roughly proportional?
Justin Florentine
so half of our planning is a surprise?
Will Corcoran
Reacted to "so half of our planning is a surprise?" with 🎯
nixo
Reacted to "so half of our planning is a surprise?" with 🎯
Kevaundray Wedderburn
Reacted to "IMG_9362.jpeg" with 😄
Will Corcoran
Replying to "so half of our plann..." Plan is just another word for guess
Ben Adams
Replying to "Is AI accurate on estimating? Sometimes AI quotes like 2 - 7 days for building something, but it actually took 20mins to finish everything. This is engineering hours btw" Nothing takes less than half a day
Parithosh Jayanthi
Reacted to "Let’s ask the LLM if we should ditch Hegota entirely and panic and ship PQ instead" with 😅
nixo
Reacted to "Plan is just another word for guess" with 😂
Mario Vega
Reacted to "so half of our planning is a surprise?" with 🎯
Mario Vega
Reacted to "Plan is just another word for guess" with 😂
potuz
Ohh boy when that chart adds the CL side
spencer
create your own fork :)
nixo
Reacted to "create your own fork :)" with 💅
Kevaundray Wedderburn
Replying to "Let’s ask the LLM if..." @Csaba are you forgetting something; what if it makes a mistake?
Justin Florentine
Next up: lets add more stuff to Hegota!
Toni Wahrstätter
Replying to "so half of our plann..." ...why it never gets boring in this space
Kevaundray Wedderburn
Reacted to "Next up: lets add more stuff to Hegota!" with 😁
Trent Van Epps
This complexity score is really interesting, curious how we can tune it to inform the fork process
Parithosh Jayanthi
Reacted to "...why it never gets boring in this space" with 😂
felipe
Reacted to ...why it never gets... with "😂"
Csaba
Replying to "Is AI accurate on es..." One of my favourite quotes from my agent (not exact, but the essence). “ You are right, my apologies, there are 365 days in a year and I forgot to multiply by that”. So no, they are no accurate.
nixo
Reacted to "One of my favourite quotes from my agent (not exact, but the essence). “ You are right, my apologies, there are 365 days in a year and I forgot to multiply by that”. So no, they are no accurate." with 😂
felipe
Reacted to One of my favourite ... with "😂"
Ben Edgington
Reacted to "One of my favourite quotes from my agent (not exact, but the essence). “ You are right, my apologies, there are 365 days in a year and I forgot to multiply by that”. So no, they are no accurate." with 😂
cayman
Reacted to "One of my favourite ..." with 😂
soispoke
A réagi à "Next up: lets add ..." avec 😁
NC
I don’t think measuring or benchmarking past EIPs would be accurate. As past EIPs related data are already used to train the LLM.
potuz
Replying to "Is AI accurate on es..." That’s because you read their outcome… as long as no-one reads what the LLM spits out they are flawless
spencer
i love that we have to use "human" now in meetings
Luca Donno | L2BEAT
Reacted to "One of my favourite quotes from my agent (not exact, but the essence). “ You are right, my apologies, there are 365 days in a year and I forgot to multiply by that”. So no, they are no accurate." with 😂
Luca Donno | L2BEAT
Reacted to "That’s because you read their outcome… as long as no-one reads what the LLM spits out they are flawless" with 👍
Pooja Ranjan
Reacted to "i love that we have to use "human" now in meetings" with 😅
Karim Taam
Replying to "i love that we have ..." In the future agent will not talk anymore about human during ACD
Csaba
Reacted to "@Csaba are you forgetting something; what if it makes a mistake?" with 😂
potuz
Reacted to "i love that we have to use "human" now in meetings" with 😅
nixo
Reacted to "i love that we have to use "human" now in meetings" with 🥩
Dustin
Well, it's also a question of which specs
Dustin
7549 had this ramifying thing
Dustin
oops bunch of beacon APIs need changes too
Csaba
Reacted to "That’s because you read their outcome… as long as no-one reads what the LLM spits out they are flawless" with 🤣
Will Corcoran
Very cool study, dan!
Dustin
so focusing too much on one spec (CL or EL or etc) is too narrow
Justin Florentine
very helpful work, thanks!
ethDreamer (Mark)
I think it would just be worth having a multidimensional score, and one of those scores could be implementation complexity in each client
Dustin
Ditto "quick slots"
Kevaundray Wedderburn
eyes
Parithosh Jayanthi
Replying to "i love that we have ..." Can we rename human to meatProxy?
Dustin
its scope/complexity is much beyond any one spec document
Trent Van Epps
Reacted to "eyes" with 👁️
spencer
Reacted to "I think it would j..." with ➕
Luca Donno | L2BEAT
Reacted to "eyes" with 👁️
Toni Wahrstätter
let's do a small fork
ignacio
Reacted to "eyes" with 👁️
Will Corcoran
Reacted to "let's do a small fork" with ➕
spencer
Reacted to "let's do a small f..." with 😅
Parithosh Jayanthi
Reacted to "eyes" with 👁️
Parithosh Jayanthi
Reacted to "let's do a small fork" with 😅
Csaba
Reacted to "Can we rename human to meatProxy?" with 👍
danceratopz
Reacted to "I think it would jus..." with 💙
Maria Silva
Replying to "let's do a small for..." I don’t think we can do a “small fork” anymore. The CFI’s are already large 😅
Will Corcoran
Replying to "let's do a small for..." Like a caviar fork
Parithosh Jayanthi
Replying to "let's do a small for..." Ah is that the mythical tech debt fork we keep talking about?
nixo
Reacted to "Can we rename human to meatProxy?" with 👍
ignacio
Reacted to "Ah is that the mythical tech debt fork we keep talking about?" with 😂
Trent Van Epps
Reacted to "🍿" with ❌
Mario Vega
Reacted to "I think it would just be worth having a multidimensional score, and one of those scores could be implementation complexity in each client" with ➕
Mario Vega
Reacted to "I think it would just be worth having a multidimensional score, and one of those scores could be implementation complexity in each client" with 💙
spencer
Reacted to "Can we rename huma..." with 👍
spencer
Reacted to "Can we rename huma..." with 😅
potuz
Reacted to "let's do a small fork" with 😂
Łukasz Rozmej
https://github.com/banteg/trace-interop
Justin Florentine
Replying to "let's do a small for..." yes, often when new capacity is unlocked it is immediately used for craftsmanship and strategic longtermism.
Csaba
We implement every idea on earth, DFId, CFId, the ones not even born ;-)
Justin Florentine
Replying to "let's do a small for..." and not short term profit. that NEVAR happens nope
Ben Adams
Replying to "We implement every idea on earth, DFId, CFId, the ones not even born ;-)" EIP Singularity
Luis Pinto
So I’m curious to know, in Nethermind are humans still reviewing code or not? Is it all fully automated?
spencer
in general i strongly agree with the sentiment, but i think we need to discuss/plan more on how to execute it all
Ben Adams
Replying to "So I’m curious to know, in Nethermind are humans still reviewing code or not? Is it all fully automated?" No; isn't automated
Ben Adams
Replying to "So I’m curious to know, in Nethermind are humans still reviewing code or not? Is it all fully automated?" We have some automation
Kevaundray Wedderburn
Can you comment on the quality of the implementation vs when the PRs were “human grown” and also the number of bugs that were fixed after the PR was merged?
NC
Replying to "So I’m curious to kn..." And do reviewers actually understand all CFI’ed EIPs
Kevaundray Wedderburn
+1 on better tooling for solving ambiguities in specs
Łukasz Rozmej
I won't be in Mumbai 😢
Will Corcoran
Reacted to "I won't be in Mumbai 😢" with 😥
FLCL (Nethermind)
Replying to "+1 on better tooling for solving ambiguities in specs" More FV
Ben Adams
Replying to "So I’m curious to know, in Nethermind are humans still reviewing code or not? Is it all fully automated?" I think we've only CFI'ed the easy ones ;)
soispoke
feels like we're already accelerating but more tooling for EIP champions would be great
potuz
And it was complex anyway 🙂
jochem-brouwer
This also means that the spec has to be correct, and that it describes what we want. AI will ship what we write but the spec is then the single source of truth. So this means we need more eyes on the spec, agents can do the implementation, but we need to have more eyes on the spec itself (also from ecosystem)
Csaba
Reacted to "I think we've only CFI'ed the easy ones ;)" with 😃
Will Corcoran
https://x.com/nero_eth/status/2108147056204222813?s=46
Will Corcoran
Replying to "https://x.com/nero_e..." PRs are accelerating
Justin Traglia
Reacted to "And it was complex anyway 🙂" with 😅
cayman
Reacted to "This also means that..." with 👍
Csaba
Reacted to "This also means that the spec has to be correct, and that it describes what we want. AI will ship what we write but the spec is then the single source of truth. So this means we need more eyes on the spec, agents can do the implementation, but we need to have more eyes on the spec itself (also from ecosystem)" with 👍
ethDreamer (Mark)
Reacted to "And it was complex anyway 🙂" with 😅
Łukasz Rozmej
Replying to "This also means that the spec has to be correct, and that it describes what we want. AI will ship what we write but the spec is then the single source of truth. So this means we need more eyes on the spec, agents can do the implementation, but we need to have more eyes on the spec itself (also from ecosystem)" AI can also review and give lot of feedback on the spec
Francesco D'Amato
Reacted to "AI can also review and give lot of feedback on the spec" with 👍
Luca Donno | L2BEAT
Reacted to "AI can also review and give lot of feedback on the spec" with 👍
Luca Donno | L2BEAT
Reacted to "This also means that the spec has to be correct, and that it describes what we want. AI will ship what we write but the spec is then the single source of truth. So this means we need more eyes on the spec, agents can do the implementation, but we need to have more eyes on the spec itself (also from ecosystem)" with 👍
CPerezz
Reacted to "AI can also review a..." with 👍
Francesco D'Amato
Replying to "This also means that..." And we can also try to state and prove as many hard properties as possible about the spec
spencer
Reacted to "AI can also review..." with 👍
Parithosh Jayanthi
Reacted to "I won't be in Mumbai 😢" with 😥
Łukasz Rozmej
Our AI reviews finds a lot more issues in code than humans ever did
danceratopz
Reacted to "feels like we're alr..." with 👍
danceratopz
Replying to "feels like we're alr..." that's the plan!
Łukasz Rozmej
That's marketing I wanted a number :)
Will Corcoran
Reacted to "feels like we're already accelerating but more tooling for EIP champions would be great" with 👍
CPerezz
Replying to "feels like we're alr..." What kind of tooling they need?
potuz
Replying to "This also means that..." @Francesco D'Amato is there an effort of trying to get the spec to be amenable for Lean and provide spec changes with Lean proofs of formal properties?
danceratopz
Replying to "feels like we're alr..." I think an automated spec checker would be helpful. Like linting, but for EIP specs.
Parithosh Jayanthi
Replying to "That's marketing I w..." 2x forks + intel tick & tok style - one big and one small
CPerezz
Replying to "feels like we're alr..." As champion. With AI you can monitor, implement specs as well as create devnets and help a lot core devs
Justin Traglia
We strive to be simple/concise/readable
Parithosh Jayanthi
Replying to "That's marketing I w..." And put the cleanup topics strictly in the tock fork
potuz
Context diversity
Enrico Del Fante
Replying to "Context diversity" And model 🙂
Csaba
Oh, an LLM can come up with a different answer each time you ask it. Client diversity solved
Łukasz Rozmej
I actually somewhat disagree spec needs to cover all edge cases - so kind of be verbose?
potuz
Reacted to "And model 🙂" with 💯
CPerezz
Replying to "feels like we're alr..." You can be the linter. If you are the champion. That’s your job essentially no? I mean, i’m not against. But a single person with AI doesn’t need a lot besides knowledge on the protocol and a bit on the clients architecture
Karim Taam
Replying to "Context diversity" Claude max vs Claude extra
Trent Van Epps
Replying to "That's marketing I w..." The “no small forks” is really culturally engrained at this point because people dont know how long they will have to wait. We would need to place some significant constraints on scope size to make that work
Gary Schulte
Replying to "This also means that..." I agree ai reviews are really good. AI PRs tend to be ruthlessly short term perspective, creating a lot of bespoke disconnected code. The review load is really really heavy, and the incentive to go faster means lower quality review
Csaba
Replying to "Oh, an LLM can come ..." Just raise the temperature :-)
danceratopz
Replying to "feels like we're alr..." Yes, but the work is not always done 🥹
danceratopz
Replying to "feels like we're alr..." I agree with you, but inconsistent specs still land.
Francesco D'Amato
@potuz I don’t know, but I think there should be one, or at least an agreed upon approach
Łukasz Rozmej
Replying to "That's marketing I wanted a number :)" But do we?
potuz
Reacted to "@potuz I don’t know, but I think there should be one, or at least an agreed upon approach" with 💯
Łukasz Rozmej
Replying to "That's marketing I wanted a number :)" We can do a lot in fork too
Łukasz Rozmej
So for example we should do I* headliner scoping soon after we finish Hegota scoping
spencer
at the PFI stage, id want ai to auto implement on all clients/eels/cl specs, its free. then a lead agent to aggregate all implementations -> to give open questions on the spec gives us meatProxy's something to start with!
CPerezz
Replying to "feels like we're alr..." Well, IIUC, we shouldn’t have CFI’d an EIP that had inconsistent specs. Client teams are supposedly looking into them to judge their marks. Thus, that should have been a lot of AI rounds already shoved into the EIP spec to either improve or discard. That’s at least what happened with SETCODEFROM
spencer
Replying to "So for example we ..." agree here, for the headliners
potuz
Replying to "This also means that..." Problem is probably the base state… but perhaps we can start adding a side spec which is in lean and we can prove some desired properties about the current spec at least
Łukasz Rozmej
exactly SOMETHING is better than the before state
Luis Pinto
I disagree that client diversity hurts specs. If each one provides its input to improve the spec in the end the spec is better than when you started. I’ve seen this happening with EIP-8037
Kevaundray Wedderburn
@lukasz I think we want high assurance that the specs are correct and diversity of specs is one way to do it. But for this to be effective, different people need to write the specs without copying each other and for efficiency it would make sense to prove they are equivalent, rather than using differential testing for example
Łukasz Rozmej
and TBH those SOMETHINGS are now really good
CPerezz
Reacted to "exactly SOMETHING is..." with 👍
ethDreamer (Mark)
Reacted to "and TBH those SOMETHINGS are now really good" with 💯
danceratopz
Reacted to "+1 on better tooling..." with 🙌
Trent Van Epps
Replying to "That's marketing I w..." I think it will come down to the harder to automate parts, as others have pointed out. Testing, speccing, scoping, testnets
danceratopz
Reacted to "This also means that..." with 👍
Łukasz Rozmej
Replying to "@lukasz I think we want high assurance that the specs are correct and diversity of specs is one way to do it. But for this to be effective, different people need to write the specs without copying each other and for efficiency it would make sense to prove they are equivalent, rather than using differential testing for example" FV the spec? :D
CPerezz
Ben is cheating…
cayman
Reacted to "exactly SOMETHING is..." with 👍
nixo
https://forkcast.org/eips/7907/
Luis Pinto
Replying to "@lukasz I think we w..." You just prove what is written, not that it solves the problem you wanted to solve
Dustin
Increasing the contract size turns out to have been, notably. much higher "complexity" than was at all obvious
Dustin
It took down almost every EL
CPerezz
Replying to "Ben is cheating…" He doesn’t want to write better SIMD code
Francesco D'Amato
Replying to "This also means that..." Not exactly what you’re getting at, but for example I found some bug in FCR due to the gloas fork-choice changes while trying to port the FCR security proofs to gloas. And that was with an ad hoc model just for FCR, with a lot of the spec outside of fork-choice just replaced by assumptions
Dustin
but the EIP was very short
spencer
Reacted to "Ben is cheating…" with 😅
Kevaundray Wedderburn
Replying to "@lukasz I think we w..." You still need some validation to check that the specs are specifying your “true intent”
Ameziane Hamlat
Supportive of the EIP
marc | wolovim
Hegota client preferences: https://forkcast.org/upgrade/hegota/client-priority/?quick=el
Łukasz Rozmej
Replying to "Increasing the contract size turns out to have been, notably. much higher "complexity" than was at all obvious" that was my gut feeling when 256kb contracts were proposed
Francesco D'Amato
Replying to "This also means that..." I think there’s a lot that we could do even with a side spec
Will Corcoran
Curious how people would feel about some degree of multi-fork planning (i.e. a couple of forks worth of headliners) for instance: moving the strawmap to the steelmap.
Justin Traglia
Reacted to "You still need some validation to check that the specs are specifying your “true intent”" with 👍
Justin Traglia
Reacted to "You just prove what is written, not that it solves the problem you wanted to solve" with 👍
Łukasz Rozmej
Replying to "Increasing the contract size turns out to have been, notably. much higher "complexity" than was at all obvious" I higly voiced not to go over 64
potuz
Reacted to "I think there’s a lot that we could do even with a side spec" with 👍
Mario Vega
Reacted to "You still need some validation to check that the specs are specifying your “true intent”" with 👍
Łukasz Rozmej
Replying to "Ben is cheating…" Not even AI can write better SIMD code than Ben
CPerezz
Reacted to "Not even AI can writ..." with 😂
Ben Adams
Reacted to "He doesn’t want to write better SIMD code" with 😅
Łukasz Rozmej
Replying to "Ben is cheating…" *Not even SI
Gary Schulte
Reacted to "Not even AI can write better SIMD code than Ben" with 🔥
Gary Schulte
Reacted to "*Not even SI" with 🤮
Ben Adams
Replying to "Ben is cheating…" I can't write a better SSD driver for all validators though ;)
Łukasz Rozmej
Replying to "Curious how people would feel about some degree of multi-fork planning (i.e. a couple of forks worth of headliners) for instance: moving the strawmap to the steelmap." I'm for for I* now
Dustin
Replying to "Increasing the con..." Even 64 proved problematic
marc | wolovim
Zoom in @nixo 🙏
CPerezz
Replying to "Curious how people w..." Partitioned Binary Tree SFI plz
Will Corcoran
Replying to "Curious how people w..." …if the decision making process / scoping process is the limiting factor… this is a viable way to overcome that
spencer
Replying to "Curious how people..." Same but for the headliners only.
Łukasz Rozmej
B-tier for us is fine in fork
Will Corcoran
Reacted to "Same but for the headliners only." with 👍
Trent Van Epps
Committing far in advance has historically been difficult as contexts/priorities/technical preferences change. Hopefully this changes if we accelerate cadence Sometimes feels like we’re still implementing stuff from 2 years ago
ethDreamer (Mark)
Replying to "Curious how people w..." Less bullish on multi-fork unless there’s a clear next priority because research on new upgrades often obsoletes the direction we previously thought we were headed
spencer
Replying to "Curious how people..." I also don't think we should rush it now, but when we get to the end of H*, we should have the headliners for J* ready, so maybe moving in that direction for the next forks
soispoke
Répondre à "Curious how people..." trade off with this is more chances that the vibe shifts and it ends up DFIing headliners
Mario Vega
Replying to "@lukasz I think we w..." This is the most important part IMO, the AI can just run with the text that is written in the EIP and miss something that would be obvious to a human core dev, we had this just happen when implementing one of hegota’s CFI’d EIPs
Iván | ethrex
+1 to what Carlos says
CPerezz
Reacted to "+1 to what Carlos sa..." with ❤️
felipe
Reacted to "This is the most i..." with 👍
Francesco D'Amato
Replying to "Curious how people w..." Once we actually get to few months forks, we can do this. Before that, it’s just going to waste a bunch of time
Lea Na
A réagi à "+1 to what Carlos ..." avec ❤️
CPerezz
Is our way to be PQ safe
CPerezz
WTF
danceratopz
Reacted to "You still need some ..." with 👍
Mario Vega
Replying to "@lukasz I think we w..." That said, we can always get better at prompting the AI
spencer
Reacted to "trade off with thi..." with 👍
spencer
Reacted to "Less bullish on mu..." with 👍
danceratopz
Reacted to "This is the most imp..." with 👍
Will Corcoran
Replying to "Curious how people w..." It would be great if you all would stop coming up with some many great ideas.
Dustin
Reacted to "Less bullish on mu..." with 👍
Danno Ferrin (shemnon)
Retire EOAs
CPerezz
Reacted to "Retire EOAs" with 👍
Barnabas
Reacted to "Retire EOAs" with 👍
Iván | ethrex
Reacted to "Retire EOAs" with 🔥
felipe
Reacted to "Less bullish on mu..." with 👍
Trent Van Epps
Replying to "@lukasz I think we w..." @Mario Vega which EIP was this?
Enrico Del Fante
Reacted to "Retire EOAs" with 👍
CPerezz
Erigon? Any takes?
Enrico Del Fante
Pls CFI both 🙂
CPerezz
Reacted to "Pls CFI both 🙂" with 👏
CPerezz
Reacted to "Pls CFI both 🙂" with ❤️
Luca Donno | L2BEAT
Reacted to "Pls CFI both 🙂" with ❤️
felipe
Reacted to "That said, we can ..." with 👍
Mario Vega
Replying to "@lukasz I think we w..." https://eips.ethereum.org/EIPS/eip-7668 , the text never mentions the Engine API, so it was automatically assumed that the full bloom should still be passed in there
Trent Van Epps
Reacted to "https://eips.ethereum.org/EIPS/eip-7668 , the text never mentions the Engine API, so it was automatically assumed that the full bloom should still be passed in there" with 👁️
Trent Van Epps
Wanted to get it into the chat transcript so future agents can learn from the mistake =)
Mario Vega
Reacted to "Wanted to get it into the chat transcript so agents can learn from the mistake =)" with 🚀
spencer
can we dfi the rest now 😅
Toni Wahrstätter
Reacted to "can we dfi the rest now 😅" with 🤘
Barnabas
Reacted to "can we dfi the rest now 😅" with 🤘
danceratopz
Reacted to "can we dfi the rest ..." with 🤘
Łukasz Rozmej
yep we bump to A
Iván | ethrex
Reacted to "yep we bump to A" with 🔥
Luca Donno | L2BEAT
Reacted to "yep we bump to A" with 🔥
Ladislaus
Reacted to "yep we bump to A" with 🔥
Kevaundray Wedderburn
Reacted to "yep we bump to A" with 🔥
CPerezz
Reacted to "yep we bump to A" with 🔥
Iván | ethrex
CFI!
Toni Wahrstätter
with frame + 2d nonces + recent roots we already get 3 new system contracts, this would be yet another one
Łukasz Rozmej
Reth is busy copying optimizations :P
Luca Donno | L2BEAT
Reacted to "Reth is busy copying optimizations :P" with 😂
Daniil Ankushin | Nethermind
Reacted to "Reth is busy copying optimizations :P" with 😂
jochem-brouwer
Antwoord verzenden naar "with frame + 2d no..." This uses eip-2935 contract :)
Ben Adams
Reacted to "Reth is busy copying optimizations :P" with 😂
danceratopz
Reacted to "in general i strongl..." with ➕
Justin Florentine
Replying to "Reth is busy copying..." Bonsai storage has entered the chat
CPerezz
Reacted to "Reth is busy copying..." with 👀
felipe
Reacted to "Retire EOAs" with 🔥
Barnabas
Reacted to "Reth is busy copying optimizations :P" with 🧂
CPerezz
Reacted to "Reth is busy copying..." with 🧂
danceratopz
Reacted to "Ah is that the mythi..." with 😂
Łukasz Rozmej
Reacted to "Reth is busy copying optimizations :P" with 🧂
Toni Wahrstätter
Replying to "with frame + 2d nonces + recent roots we already get 3 new system contracts, this would be yet another one" ahh then I understood it wrong. This is then basically a no-op?
Luis Pinto
Reacted to "Bonsai storage has entered the chat" with 🙀
Kevaundray Wedderburn
Replying to "with frame + 2d nonc..." In some sense, this is more a simplification because we store the block hashes in state but are not using it
Francesco D'Amato
Replying to "with frame + 2d nonc..." It increases the cost a lot no?
Justin Traglia
Replying to "Reth is busy copying..." Does it really matter? It’s all open source. Let good ideas propagate.
Maria Silva
Replying to "with frame + 2d nonc..." There is the pricing
danceratopz
Reacted to "let's do a small for..." with 🤏
Justin Florentine
Reacted to "Does it really matter? It’s all open source. Let good ideas propagate." with ➕
Luis Pinto
Reacted to "Does it really matter? It’s all open source. Let good ideas propagate." with 💯
Luca Donno | L2BEAT
Reacted to "This uses eip-2935 contract :)" with 👍
Kevaundray Wedderburn
Replying to "with frame + 2d nonc..." Yeah its pricing changes
felipe
Reacted to "Does it really mat..." with 💯
Luca Donno | L2BEAT
Reacted to "It increases the cost a lot no?" with 👍
Justin Florentine
Replying to "Reth is busy copying..." if we don't steal from each other our agents will
danceratopz
Reacted to "Human evals also vis..." with 💙
Karim Taam
Replying to "Reth is busy copying..." We should just try to not have the same client multiple times and keep diversity in the code
Francesco D'Amato
Replying to "with frame + 2d nonc..." Can we not read it from the contract to remove the special case in the state transition, but keep the pricing as is, and just mandate clients to load that state?
jochem-brouwer
Heeft gereageerd op "We should just try..." met 👍
felipe
Reacted to "We should just try..." with 👍
Joss
Reacted to "Does it really mat..." with 💯
Francesco D'Amato
Replying to "with frame + 2d nonc..." I guess it adds pricing special cases…
Luis Pinto
Reacted to "We should just try to not have the same client multiple times and keep diversity in the code" with 👍
Toni Wahrstätter
Replying to "with frame + 2d nonces + recent roots we already get 3 new system contracts, this would be yet another one" what it changes is basically adding traces to BLOCKHASH.
Maria Silva
Replying to "with frame + 2d nonc..." I am not sure what you mean, @Francesco D'Amato
Luca Donno | L2BEAT
Reacted to "We should just try to not have the same client multiple times and keep diversity in the code" with 👍
Francesco D'Amato
Replying to "with frame + 2d nonc..." As it is, it will get way more expensive even if a bunch of txs in the block read it, right?
potuz
Replying to "@lukasz I think we w..." FV in the specs is the right way to go, it’s not “to prove what is written” blindly, but rather asserting properties that we want from the spec eg. “finalization does not revert”, and formally proving that the spec satisfies it
Justin Florentine
Replying to "Reth is busy copying..." use-case diversity is our friend: sometimes we do things in a specific way for tradfi/private networks. those are more durable, diverse choices
Maria Silva
Replying to "with frame + 2d nonc..." Right, we need to price it based on worst-case, as we are doing with all the other opcodes
Francesco D'Amato
Replying to "with frame + 2d nonc..." Well, in practice clients have those values anyway
Łukasz Rozmej
Reacted to "We should just try to not have the same client multiple times and keep diversity in the code" with 👍
Francesco D'Amato
Replying to "with frame + 2d nonc..." So I wouldn’t call it worst case
Łukasz Rozmej
Replying to "Reth is busy copying optimizations :P" @Karim Taam that's what I said: https://x.com/URozmej/status/2107894718965452847
Justin Florentine
VOPS
Kevaundray Wedderburn
Replying to "with frame + 2d nonc..." I think its worse case based on a client that will do the simple thing and read the contract (ie sload cost)?
jonny (geth)
Replying to "Reth is busy copying..." please apologize for copying him
Francesco D'Amato
Replying to "with frame + 2d nonc..." Yeah I guess it’s not clear to me why it is necessary to assume this behaviour, instead of (for example) mandating that clients load the values before processing the block
Maria Silva
Replying to "with frame + 2d nonc..." Sorry, I was missing this thread 😅 Yes, we need to assume clients follow the spec (i.e., go to the state DB) and we need to price by it
Francesco D'Amato
Which would have the equivalent behaviour as today but without treating it as “special values outside of the state”
Francesco D'Amato
Replying to "with frame + 2d nonc..." I don’t agree that “we need to"
Francesco D'Amato
Replying to "with frame + 2d nonc..." Or at least I don’t see why
Justin Florentine
hash based sigs are gonna be on the evm
Karim Taam
I think we always end up with similarities in the code, because there aren't a thousand ways to optimize certain things. But we just have to remain vigilant, and even more with AI that will create in the future one clients with different languages
Francesco D'Amato
Replying to "with frame + 2d nonc..." (That said, I don’t know if this becoming much more expensive would be a problem for some use cases, if it’s not then whatever)
CPerezz
Didn’t @Kevaundray Wedderburn already check usage on this opcode and turned out almost no-one uses it? Can’t we just get rid of this instead?
soispoke
going great
Iván | ethrex
Reacted to "going great" with 😅
nixo
Reacted to "Screenshot 2026-10-08 at 16.07.15.png" with 😅
Parithosh Jayanthi
Reacted to "going great" with 😅
Parithosh Jayanthi
Reacted to "Screenshot 2026-10-08 at 16.07.15.png" with 😅
Łukasz Rozmej
Replying to "Screenshot 2026-10-08 at 16.07.15.png" but this doesn't take AI into account
Łukasz Rozmej
Replying to "Screenshot 2026-10-08 at 16.07.15.png" which can accelerate
Luca Donno | L2BEAT
Reacted to "going great" with 😅
felipe
Reacted to "Screenshot 2026-10..." with 😅
Parithosh Jayanthi
Replying to "Screenshot 2026-10-08 at 16.07.15.png" E = mc2 + AI vibes
felipe
Reacted to "going great" with 😅
spencer
Reacted to "going great" with 😅
Luca Donno | L2BEAT
Reacted to "E = mc2 + AI vibes" with 😂
spencer
Reacted to "Screenshot 2026-10..." with 😅
Iván | ethrex
We could do both 😁
Barnabas
Replying to "Screenshot 2026-10-08 at 16.07.15.png" is AI the limit today?
Barnabas
Replying to "Screenshot 2026-10-08 at 16.07.15.png" I feel like we are already non AI limited today
Barnabas
Replying to "Screenshot 2026-10-08 at 16.07.15.png" but rather human limited
Justin Florentine
it gets harder to not consider size/scope when you get down to the bottom of the list.
Łukasz Rozmej
on the other hand this is so low value, that we already spent too much time discussing it
Iván | ethrex
Reacted to "on the other hand th..." with 😅
Justin Florentine
the danger of arguing past each other increases
spencer
Reacted to "E = mc2 + AI vibes" with 😂
Łukasz Rozmej
implementing it would be faster as well as DFI
Luca Donno | L2BEAT
Reacted to "on the other hand this is so low value, that we already spent too much time discussing it" with 👍
Ben Adams
Reacted to "Ben is cheating…" with 😅
soispoke
A réagi à "E = mc2 + AI vibes" avec 😂
Iván | ethrex
Reacted to "implementing it woul..." with ☝️
ignacio
Reacted to "implementing it would be faster as well as DFI" with ☝️
Csaba
Reacted to "E = mc2 + AI vibes" with 😂
Parithosh Jayanthi
Well DFI is less code debt 😄
potuz
Reacted to "E = mc2 + AI vibes" with 😂
spencer
Reacted to "Well DFI is less c..." with 😅
jochem-brouwer
for this eip also consider this comment: https://ethereum-magicians.org/t/eip-7666-evm-ify-the-identity-precompile/19445/2?u=jochem-brouwer
Łukasz Rozmej
Reacted to "for this eip also consider this comment: https://ethereum-magicians.org/t/eip-7666-evm-ify-the-identity-precompile/19445/2?u=jochem-brouwer" with 👍
lightclient
why not dfi?
Łukasz Rozmej
Replying to "for this eip also consider this comment: https://ethereum-magicians.org/t/eip-7666-evm-ify-the-identity-precompile/19445/2?u=jochem-brouwer" DFI then
CPerezz
Replying to "why not dfi?" Why not removing the opcode?
lightclient
DFI! DFI! DFI!
Justin Traglia
Looks like a lot of CFI’d EIPs 😅
Mario Vega
Reacted to "E = mc2 + AI vibes" with 😂
Will Corcoran
Reacted to "Looks like a lot of CFI’d EIPs 😅" with 🫠
soispoke
A réagi à "Screenshot2026_10_..." avec 🚢
Barnabas
Replying to "Looks like a lot of ..." ppl just want everything
nixo
Reacted to "ppl just want everything" with 💯
Łukasz Rozmej
Is ML-DSA is secure enough not to be broken in near future (Vitalik post)?
Justin Traglia
Reacted to "Looks like a lot of CFI’d EIPs 😅" with 🫠
spencer
Replying to "Is ML-DSA is secur..." yeah i feel dfi
Antonio Sanso
Replying to "Is ML-DSA is secure ..." Unclear right now
potuz
Are we OK with lattices? The AI overlords haven’t had time to attack them yet
Justin Traglia
Reacted to "Are we OK with lattices? The AI overlords haven’t had time to attack them yet" with 👎
Łukasz Rozmej
Replying to "Is ML-DSA is secure enough not to be broken in near future (Vitalik post)?" DFI and wait for another fork?
Vitalik
Replying to "why not dfi?" evm-ifying is strictly less backwards compat-breaking than removing
Justin Traglia
Replying to "Are we OK with latti..." No.
Csaba
Replying to "Is ML-DSA is secure ..." Isn’t there some industry-wide standardisation effort behind it?
Justin Florentine
tradfi is converging on this
Luca Donno | L2BEAT
Reacted to "Looks like a lot of CFI’d EIPs 😅" with 🫠
Luca Donno | L2BEAT
Reacted to "Are we OK with lattices? The AI overlords haven’t had time to attack them yet" with 👎
felipe
Reacted to "Screenshot2026_10_..." with 🚢
Francesco D'Amato
Reacted to "FV in the specs is the right way to go, it’s not “to prove what is written” blindly, but rather asserting properties that we want from the spec eg. “finalization does not revert”, and formally proving that the spec satisfies it" with 👍
Luca Donno | L2BEAT
Reacted to "yeah i feel dfi" with 👍
Csaba
Replying to "Is ML-DSA is secure ..." I just recall something, apologies if not.
CPerezz
fair
Dustin
what what, "people need choices" is the whole rationale?
danceratopz
Replying to "feels like we're alr..." That should be the case, yes! The friendly pitch is making tooling to help EIP champions. The less friendly pitch is making a public EIP Spec Health dashboard to keep authors accountable.
lightclient
my feeling on ML-DSA is that is the official recommendation for many countries and we already made the mistake of waiting a very long time to have secp256r1 which was also the industry standard
Vitalik
hash-based sigs are provably reducible to hash security, if hashes are not secure, ethereum is completely broken
CPerezz
If we go for SETCODEFROM, we can on the meantime use solidity-optimized impls of whatever we want which are close to 1M or similar gas. And with the gas increases, this is not an issue. That leaves anyone the choice of what to use. And we don’t overcommit early on any scheme
jochem-brouwer
Heeft gereageerd op "hash-based sigs ar..." met 👍
Iván | ethrex
Reacted to "hash-based sigs are ..." with 👍
Justin Traglia
Reacted to "hash-based sigs are provably reducible to hash security, if hashes are not secure, ethereum is completely broken" with 👍
Parithosh Jayanthi
Reacted to "Looks like a lot of CFI’d EIPs 😅" with 🫠
lightclient
Reacted to "If we go for SETCODEFROM, we can on the meantime use solidity-optimized impls of whatever we want which are close to 1M or similar gas. And with the gas increases, this is not an issue. That leaves anyone the choice of what to use. And we don’t overcommit early on any scheme" with 👍
spencer
Reacted to "If we go for SETCO..." with 🔥
jochem-brouwer
Heeft gereageerd op "If we go for SETCO..." met 👍
soispoke
A réagi à "If we go for SETCO..." avec 👍
lightclient
Reacted to "ppl just want everything" with 💯
Mario Vega
Reacted to "feels like we're already accelerating but more tooling for EIP champions would be great" with 👍
potuz
Reacted to "hash-based sigs are provably reducible to hash security, if hashes are not secure, ethereum is completely broken" with 👍
danceratopz
Reacted to "soispoke vous a envo..." with 😅
Benedikt Wagner
Reacted to "hash-based sigs are provably reducible to hash security, if hashes are not secure, ethereum is completely broken" with 👍
Matteo Vena | Riva Labs
Reacted to "hash-based sigs are provably reducible to hash security, if hashes are not secure, ethereum is completely broken" with 👍
Justin Traglia
Reacted to "ppl just want everything" with 💯
CPerezz
Reacted to "That should be the c..." with 👍
ignacio
Looking at recently-sih added 7 BLS precompiles with bleeding eyes...
CPerezz
Reacted to "Looking at recently-..." with 😂
Maria Silva
Reacted to "Looking at recently-sih added 7 BLS precompiles with bleeding eyes..." with 😂
Ansgar Dietrichs
I agree it’s worth exploring whether we can do something without shipping the actual precompile, like what Danno just said
Will Corcoran
Reacted to "I agree it’s worth exploring whether we can do something without shipping the actual precompile, like what Danno just said" with 👍
CPerezz
Replying to "Looking at recently-..." Ai companies trolled Antonio a bit here..
Iván | ethrex
Our concern with 8355 is that it is a premature decision whose cost is permanent Once a precompile is added, it is very unlikely to ever be removed There’s a whole effort moving towards a zkEVM, and it is not clear yet whether this is aligned with it, or at what cost.
CPerezz
Reacted to "I agree it’s worth e..." with 👆
Matteo Vena | Riva Labs
Reacted to "Our concern with 8355 is that it is a premature decision whose cost is permanent Once a precompile is added, it is very unlikely to ever be removed There’s a whole effort moving towards a zkEVM, and it is not clear yet whether this is aligned with it, or at what cost." with 👍🏻
Greg K | Lido
Reacted to "Our concern with 8355 is that it is a premature decision whose cost is permanent Once a precompile is added, it is very unlikely to ever be removed There’s a whole effort moving towards a zkEVM, and it is not clear yet whether this is aligned with it, or at what cost." with 👍🏻
Mario Vega
Reacted to "hash-based sigs are provably reducible to hash security, if hashes are not secure, ethereum is completely broken" with 👍
ignacio
Reacted to "Ai companies trolled Antonio a bit here.." with 😄
Vitalik
If the argument is choice, then that feels like an argument for doing one of two other things: full-on NVMPY (SIMD for ethereum), so we support all lattices including FHE, and it also would reduce gas cost of non-officially-supported hashes accelerate full 8288, "if you want something that's not SPHINCS, STARK-wrap it"
danceratopz
Reacted to "Looks like a lot of ..." with 🫠
danceratopz
Reacted to "Looks like a lot of ..." with ➕
jochem-brouwer
Heeft gereageerd op "Looking at recentl..." met 🥲
ignacio
Reacted to "Our concern with 8355 is that it is a premature decision whose cost is permanent Once a precompile is added, it is very unlikely to ever be removed There’s a whole effort moving towards a zkEVM, and it is not clear yet whether this is aligned with it, or at what cost." with 👍🏻
Justin Florentine
PQ vs. Decision Speed yall
Benedikt Wagner
Reacted to "If the argument is choice, then that feels like an argument for doing one of two other things: full-on NVMPY (SIMD for ethereum), so we support all lattices including FHE, and it also would reduce gas cost of non-officially-supported hashes accelerate full 8288, "if you want something that's not SPHINCS, STARK-wrap it"" with 👍
Greg K | Lido
Replying to "Looking at recently-sih added 7 BLS precompiles with bleeding eyes..." Useful in 2022, shipped in 2025
Łukasz Rozmej
Reacted to "hash-based sigs are provably reducible to hash security, if hashes are not secure, ethereum is completely broken" with 👍
Łukasz Rozmej
this is technical debt clearing
Matteo Vena | Riva Labs
Reacted to "If the argument is choice, then that feels like an argument for doing one of two other things: full-on NVMPY (SIMD for ethereum), so we support all lattices including FHE, and it also would reduce gas cost of non-officially-supported hashes accelerate full 8288, "if you want something that's not SPHINCS, STARK-wrap it"" with 👍
ignacio
Reacted to "Useful in 2022, shipped in 2025" with 🥲
Luis Pinto
It was lower priority I think that’s why it was C tier?
marc | wolovim
Already CFIed (7668 remove bloom filters)
Ansgar Dietrichs
I think we should have as much “PQ clarity” as possiblle by the time the fork rolls out btw. E.g. if we indeed have a full agreed ML-DSA precompile spec and reserved address by then even without shipping it, users adopting frame txs could choose to register ML-DSA keys as fallback with a verification logic that switches between precompile and native implementation, depending on availability
Justin Florentine
the juice was not worth the squeeze for Besu
Maria Silva
I think it was because the impact was fairly small
Ansgar Dietrichs
Replying to "I think we should ha..." and/or similarly for any other PQ scheme we might plan to ship in a later fork
danceratopz
Replying to "soispoke vous a envo..." @Łukasz Rozmej this version's for you (and @spencer, who suggested it!!!) https://danceratopz.github.io/retrospective-eip-complexity/results/predicted-vs-observed/?ai=0.20 [Full message cannot be displayed on this version]
Antonio Sanso
Replying to "I think we should ha..." Why if we have a sphincs working solution already ?
CPerezz
Replying to "I think we should ha..." If you have a way to use whatever scheme with SETCODEFROM and you can invalidate your ECDSA key with 8151. What more certainty you need for H*? You can be safe (in EL) if you want to pay the price.
Gianluca Felli
Ha reagito a "If the argument is..." con 👍
Barnabas
time to call Peter?
nixo
Reacted to "time to call Peter?" with 😄
Daniel Knopik
Reacted to "time to call Peter?" with 😄
CPerezz
Replying to "I think we should ha..." This at least gives us breathing room to not make early commitments that turn out to be mistakes.
Vitalik
Replying to "@lukasz I think we..." at that point "spec" starts to have two distinct meanings: human language description of what we want formal lean4 spec
Łukasz Rozmej
Reacted to "time to call Peter?" with 😄
Vitalik
Replying to "@lukasz I think we..." doing (2) multiple times and proving equivalence/compatibility between them seems like a reasonable thing to do
Alessandro
Reacted to "hash-based sigs are provably reducible to hash security, if hashes are not secure, ethereum is completely broken" with 👍
Vitalik
yeah basically the downside is that it adds the complexity that some actions are in the journal that gets reverted and others are not
Maria Silva
Replying to "Screenshot 2026-10-08 at 16.07.15.png" Why is Cancun not moving?
Vitalik
I don't think there's another good reason
jochem-brouwer
it is more fair gas journaling i would say. I could imagine that even from reverted frames you can then figure out if a slot was touched or not via some obscure gas calc logic, which might have been the reason for eip-2929 not carrying warm address/slots on reverted frames
danceratopz
Replying to "soispoke vous a envo..." Thx for the bug report
Kevaundray Wedderburn
Replying to "I think we should ha..." I think there will be other chains that use ml-dsa so standards for it would be good, even if its not the default on L1
Maria Silva
Replying to "Screenshot 2026-10-08 at 16.07.15.png" Actually only prague, Amsterdam and the hegota line are moving
danceratopz
Replying to "soispoke vous a envo..." Or was AI so much worse maybe? (Shanghai not moving either)
CPerezz
Replying to "I think we should ha..." @Antonio Sanso is leanSPHINCS fully final? Or are there any pending tasks? Could we ship tomorrow and be sure it just works?
Justin Florentine
great work team
Antonio Sanso
Replying to "I think we should ha..." Give a look to https://daisugi.fyi/ we are deploying v 0,3 this week with signature aggregation (Thx Nethermind and Ethrex for helping out)
Pooja Ranjan
https://github.com/ethereum/pm/issues/2234#issuecomment-6060792254 https://github.com/ethereum/pm/pull/2266 I’d appreciate hearing from devs on the initial version of the “Glamsterdam Ecosystem Readiness Checklist”- here or async. If it looks fine, we can get the PR merged and share the link with ecosystem stakeholders.
jochem-brouwer
next acde coffee corner time
Łukasz Rozmej
I wish ACDC had this tempo
Kevaundray Wedderburn
Reacted to "doing (2) multiple times and proving equivalence/compatibility between them seems like a reasonable thing to do" with 👍
Csaba
Reacted to "Wanted to get it into the chat transcript so future agents can learn from the mistake =)" with 😃
Greg K | Lido
Reacted to "I wish ACDC had this tempo" with 👍
Greg K | Lido
Reacted to "I wish ACDC had this tempo" with 💯
Antonio Sanso
Replying to "I think we should ha..." We cannot ship tomorrow but same as for the CL solution
Maria Silva
So, are we done with scoping?
Łukasz Rozmej
Replying to "So, are we done with scoping?" almost
nixo
Replying to "So, are we done with..." minus 7907, yes!
CPerezz
Replying to "I think we should ha..." Then SETCODEFROM + 8151 is not an option. It’s a must
Maria Silva
Replying to "So, are we done with..." Uau, that was fast!
Luca Donno | L2BEAT
Reacted to "minus 7907, yes!" with 🎉
Łukasz Rozmej
so in accelerate mindset maybe I* headliner discussion? :D
Luca Donno | L2BEAT
Reacted to "so in accelerate mindset maybe I* headliner discussion? :D" with 😃
jochem-brouwer
Heeft gereageerd op "Uau, that was fast!" met 🏎️
Barnabas
Reacted to "so in accelerate mindset maybe I* headliner discussion? :D" with 😃
Maria Silva
Reacted to "so in accelerate mindset maybe I* headliner discussion? :D" with 😃
marc | wolovim
🙏 CFI everything is fast too 🙂
CPerezz
Replying to "so in accelerate min..." PBT. What’s there to discuss? And LeanSPHINCHS with it
Barnabas
Replying to "so in accelerate min..." I star name?
danceratopz
Replying to "soispoke vous a envo..." Will add an explainer!
spencer
Reacted to "Uau, that was fast!" with 🏎️

Call summary

Targets

•Hoodi Glamsterdam fork — Monday October 26, 2026 at 17:42 UTC - 00:21:44
•Client releases for Hoodi fork — by Friday October 16, 2026 - 00:22:52

Decisions

•Hoodi fork: Monday October 26, 2026 at 17:42 UTC; client releases by October 16 - 00:21:44
•Gas limit change activates on fork epoch for Hoodi and Mainnet - 00:30:13
•EIP-7907 (contract size metering) added to PFI list; decision deferred to next call - 00:59:27
•CFI for Hegota: EIP-5920 (PAY opcode), EIP-8151 (ecRecover), EIP-8298 (SETCODEFROM), EIP-8360 (TCREATE), EIP-4758 (Deactivate SELFDESTRUCT), EIP-7709 (BLOCKHASH from storage), EIP-8077 (eth/xx networking), EIP-8116 (replace cumulative receipt fields) - 01:03:00
•DFI for Hegota: EIP-7666 (EVM-ify identity precompile), EIP-8355 (ML-DSA precompile), EIP-8374 (persist warm access sets across reverts) - 01:27:03

Highlights

Eip Proposals Hegota:
- ·EIP-7907 (contract size metering): PFI raised by Ben Adams; EIP-7954 raised contract limit without metering, creating DoS surface; decision deferred to next call - 00:59:01
- ·CFI: EIP-5920 (PAY opcode) — medium complexity, broad support - 01:03:00
- ·CFI: EIP-8151 (ecRecover deprecation) and EIP-8298 (SETCODEFROM) — paired as PQ EOA migration path - 01:06:49
- ·CFI: EIP-8360 (TCREATE opcode) and EIP-4758 (Deactivate SELFDESTRUCT) — paired; fix for issue introduced in Glamsterdam - 01:07:23
- ·CFI: EIP-7709 (read BLOCKHASH from storage) — upgraded to A-tier by Nethermind; benefits BAL light-client use case - 01:12:28
- ·CFI: EIP-8077 (eth/xx mempool networking) — conditional on deeper analysis including frame/fee fields - 01:18:39
- ·DFI: EIP-7666 (EVM-ify identity precompile) — gas pricing edge cases and low value; consensus shifted to DFI after jochem-brouwer raised spec issues - 01:27:03
- ·DFI: EIP-8355 (ML-DSA precompile) — no consensus on lattice schemes; zkEVM cost concerns; hash-based approach preferred - 01:34:02
- ·CFI: EIP-8116 (replace cumulative receipt fields) — low priority, placed at bottom of CFI list - 01:36:10
- ·DFI: EIP-8374 (persist warm access sets across reverts) — lukewarm support, deep fork boundary change not worth complexity - 01:40:31
Complexity And Tooling:
- ·danceratopz presented LLM-based retrospective EIP complexity assessments; 14 CFI/SFI EIPs already make Hegota the most complex fork on record before any additional PFIs - 00:33:49
- ·Łukasz Rozmej: Nethermind has 150+ Hegota PRs; bottleneck is coordination/testing, not implementation; advocates for better tooling and EIP champion involvement - 00:47:00
Fork Status And Schedule:
- ·Glamsterdam fork on Sepolia successful; boring fork, gas limit ramped to 200M post-fork - 00:17:24
- ·ePBS builder participation on Sepolia low: ~4 active builders (Titan/Nico/Sigma Prime), large players absent - 00:18:46
- ·Hoodi fork scheduled Monday October 26 at 17:42 UTC; client releases required by October 16 (10-day notice) - 00:21:44
- ·Gas limit change confirmed to activate on the fork epoch for both Hoodi and Mainnet - 00:30:13
- ·Mainnet shadow fork accidentally activated Gloas; now being used for Snap 2 testing with large blocks; benchmarking shadow fork relaunching next week - 00:30:38

Action Items

•Csaba (EIP-8077 champion) - EIP-8077 champion to expand proposal to include fee/frame fields beyond sender+nonce for mempool filtering - 01:18:39
•EL client teams - Review Glamsterdam Ecosystem Readiness Checklist PR (#2266) for enterprise Sepolia feedback collection - 01:41:07

Key decisions

Hoodi fork: Monday October 26, 2026 at 17:42 UTC; client releases by October 16
Scheduling of the Hoodi fork on Sepolia testnet.
Gas limit change activates on fork epoch for Hoodi and Mainnet
Confirmation of gas limit ramp activation timing.
EIP-7907 (contract size metering) added to PFI list; decision deferred to next call
PFIAdded to Proposed for Inclusion list; final decision deferred.
CFI for Hegota: EIP-5920 (PAY opcode), EIP-8151 (ecRecover), EIP-8298 (SETCODEFROM), EIP-8360 (TCREATE), EIP-4758 (Deactivate SELFDESTRUCT), EIP-7709 (BLOCKHASH from storage), EIP-8077 (eth/xx networking), EIP-8116 (replace cumulative receipt fields)
CFIConsidered for Inclusion in the Hegota upgrade.
DFI for Hegota: EIP-7666 (EVM-ify identity precompile), EIP-8355 (ML-DSA precompile), EIP-8374 (persist warm access sets across reverts)
DFIDeclined for Inclusion in the Hegota upgrade due to complexity or lack of consensus.

EIPs discussed

EIP-4758 EIP-5920 EIP-7666 EIP-7709 EIP-7907 EIP-8077 EIP-8116 EIP-8151 EIP-8298 EIP-8355 EIP-8360 EIP-8374

Links

AI Disclaimer: Some content or metadata on EIPsInsight may be AI-inferred or automatically compiled. If you find any discrepancy, please contact us at dev@avarch.org.