Two out of the three top multi-party blocks by value uplift landed in the last 2 days. Don't expect these to stay on the leaderboard for long. New significant performance improvements are already in testing. These will allow MPBC to support even larger blocks.

Here's the multi-party block with the highest gas uplift to date. MPBC increased gas usage by 1346%, adding 24.5M gas. It's Slot: 15213880 The highest number of transactions added to date is 810. Assuming users would have waited ~12s for the next slot, it saved them 2.7h. You can find it on the dashboard (Slot: 15180396).

All top blocks by gas added are by newer and smaller base builders. This includes 24.5m gas added to one block (+1346%)! MPBC rewards providing faster and more predictable inclusion to users. Blocks from smaller builders typically have comparatively more space available. Check out the MPBC dashboard to view every single MPBC block.

Fantastic to have Quasar as part of the Blockspace Forum and MPBC! Onwards & upwards..

The Swiss housing price index, annualized Sharpe of ~1.9.

Falco for Austria:

The current AI discourse reminds me of the classic 2013 "Bitcoin Citadels" story.

> Most painful DeFi Experience.

Interesting chart on quant shop AI use vs. trade duration. From an FT article - linked in replies - that is generally worth reading.

4/ Access just about any frontier model privately via Venice. Includes Opus 4.5, GPT-5.2, Gemini 3...

3/ Supports any number of parallel conversations. - Ask models to sanity check each other. - Ask multiple models to check your work. - Ask models to take on specialist roles.

2/ Chat with LLMs through the sidebar. Models can directly modify spreadsheets. Select cells to share them with the model. Everything else stays private.

🧵1/ Native Claude support for Linux spreadsheets Built a native extension that allows Claude and other frontier models to plug into calc spreadsheets. In difference to Claude Excel, it allows chatting with multiple frontier models in parallel. Privacy support via @AskVenice

The Anthropic take home was good fun. Managed to squeeze under 1.8k cycles before dinner, temporarily hovering around some folks more familiar with performance computing. My main learning is that AI agents should be given a clear function to maximize. Most valuable as a way of trying many things quickly in such a framework, rather than for cognition. A second learning is that trimming complexity and bloat regularly helps avoid dead-ends.

We're now merging blobs, up to doubling their count in underutilized blocks.

Graphs I enjoy seeing.

Palermo BA is one of the world’s great neighbourhoods.

Finally managed to get my hands on the full set 2007-2008.

Also this is nicely done -

As it happens, I’m currently reading Mozart’s letters. He, quote, is: “…most surprised and delighted about loud shouts of ‘Bravo’” in the middle of a Piano performance. Clapping during a concert is indeed traditional / appropriate. Source ->

Presented Relay Inclusion Lists (rILs) in the FOCIL breakout call today. rILs improve Ethereum's censorship resistance fast and simply. It's doable today. Recording & research post: - piped.video/watch?v=0clvpSfi… - ethresear.ch/t/relay-inclusi… Merci for the discussion @soispoke @weboftrees @jih2nn..!

SIMD-228 is a net plus for Solana, albeit the new inflation curve appears a tad aggressive. We expect the key risk to be reward variance for low-stake validators during low-volume periods. To illustrate this point in just two graphs: (1) Market APR since Jan 1st. with outlier events (e.g. TRUMP launch). (2) Market APR since Jan 1st without outlier events. We recommend voting "yes", but will split our vote to fully reflect delegator preferences. Read more here: chorus.one/articles/an-analy… Props to @un_data_wiz and @0xrafal for the excellent analysis. Thanks to @kankanivishal and @maxresnick for the discussions.

We just published a practical perspective on inclusion preconf pricing. -> Comes with a dashboard; play through different models. -> Outperforms standard heuristics significantly. Dashboard: preconfs.chorus.one Article: ethresear.ch/t/pricing-trans… Props to @benrobaglia @un_data_wiz

We'll do our best to roll our @berachain testnet performance over to mainnet, but no promises.

After cancelling my initial flight, @lufthansa has now waitlisted me for the third (!) alternate over the span of two days. There are probably ~15 ppl in the same situation. Epically bad performance for passengers and the ground staff forced to deal with the fallout.

Yeah, makes sense. For us one specific concern was that timing games increase burn, creating an unfavourable tail that again disproportionately hits solo stakers because less draws

Started to explore / build out infra for our @berachain PoL algo. Drafted up a toy dashboard to visualize things - check it out. Early Observations - -> Infra seems performant. Can query the RPC via multiple threads & it stays responsive. -> BEX is a fork of crocswap, now @ambient_finance, TIL! That’s nice. -> Still early. Eg seeing signs of what may be user agent tests in the LP-token naming. Maybe you can spot it in the image. Dashboard is not final in any way, just play at this point. It runs a live fetch, screenshot is a snapshot after startup. -->Overall, early days still, but feels exciting.

Cool talk and result! - Self-built blocks do not feature private order flow and artificially depress the base fee. - These are typically followed by mev-boost “recovery blocks” that subsequently significantly raise the gas fee. -> Decoupling between the base fee & spot demand for blockspace if vanilla & mev-boost blocks are mixed. -> Divergence expected to increase as the share of private order flow increases. -> Another reason is that builders have more time to provide a block vs. a typical vanilla proposal. Keen to read the study. A weaker result would also extend to proposer timing games, where builders have more transactions available as the slot progresses. Graph from our timing games study below.

4/ Relay-wise, no surprises. @titanbuilderxyz continues to outperform.

3/ Keeping idiosyncratic risk low is important. For example, consider querying a relay multiple times to lock-in cancelled bids. There's a solid chance the relay will rate limit. Even if it doesn't currently, it can throw up a limit at any point. Rotating IPs to bypass such a limit is not a desirable point on the risk <> reward curve. Consider the curve max. bid / current bid - does that look like cancelled bids are a sufficiently large opp. to run the additional risk of swapping out IPs? Likely not.

2/ We're outperforming our original estimate by a lot. Timing games are mostly an optimization over the hardware, networking, kernel layer. The mev-boost fork itself is generally downstream. Though some moated, non-obvious additions to mev-boost are possible: nitter.net/ChorusOne/status/17882…

1/ Our view of the mev-boost auction (v.01) -> adagio.chorus.one/ Tracks the PBS auction to simulate the payoffs with- and without timing games for our setup. The long-term upside settles around ~10%, which is a comfy risk-adjusted spot. Observations below ->

Spent the last 3 days talking preconfirms at @ZuBerlinCity. Glad we proposed some of the first-ever preconfs. Very cool..! But mostly excited about what's still cooking.. 🕶️

Bullish @symbioticfi Felix is a great guy and true OG..:

If I catch your drift correctly- These last 100ms are likely worth a maximum if arbitrageurs know they're incoming, due to inventory risk. If arbitrageurs don't know, then I'd say value tends to cluster at the expected block submission time. Attaching some empirics from our timing games study: chorus.one/reports-research/…

1/ Some quick alpha on the timing games & Solana discussion. We ran the numbers on node geographic distribution in ~Feb. Results did indicate that there was concentration around ~BeNeLux. Specifically, in one NL datacenter. Nodes in this datacenter seemed to do well -> but..

@titanbuilderxyz relay with beastly performance last week.

Live dashboard showcasing our in-validator view of the mev-boost auction. Compares our Adagio mev-boost fork to a standard setup. Landing soon alongside the methodology. Will cover longer time-horizons. Last week was a very strong week.

The start of an exciting experiment..

4/ The short answer was - yes, but due to variance, a reliable upside estimate would take a while to manifest. The graph shows the way the current bid evolves as a fraction of the max. bid over time. Taken from our study on timing games (as is the first graph) - linked below.

1/13 Adagio, our mev-boost fork, has been publicly available for 100 days now. We have been testing it on mainnet since early 2023. A quick thread on timing games and how we found them, rational competition, and resilient decentralization 🧵

Excited to present our work on validator timing games & externalities at ETH Denver today: 5pm, Neptune Stage. For a sneak peek, check out @umb_nat's excellent presentation from a recent Lido community call: piped.video/watch?v=J_N13erD…

Great panel on MEV, LVR, and all things incentive alignment with @0xMagmar @tkstanczak @ThogardPvP @maxresnick today. Thanks organizing @a41_allforone - cool event!