Hand-painted dispatches from the Layer-2 frontier — rollups, sequencers, bridges, and the tides moving between them.
Loose pages from the desk this fortnight — what caught the eye, what surprised, and what is probably nothing.
Everyone expected fragmentation by now. Instead the top three rollups are settling into polite coexistence — sharing standards, sharing dev tools, occasionally sharing users. The real war moved upstream, into shared sequencing layers nobody is naming yet.
market share, top-3 sequencers, 12mo
Aggregate exploit losses are down 71% YoY. Not because bridges got bulletproof — they got boring. Standardized primitives, shared audits, fewer bespoke contracts. Boring is good.
Smart-account adoption crossed 18% of active L2 wallets this month. The interfaces stopped explaining themselves, which is how you know it's working — people just use them.
cf. Entry 12, Q4 forecast was 9%
Where the mempool was the dirty little secret of L1 economics, DA is now where the rollup margin actually lives. Watching rollups choose between blob posts, dedicated DA layers, and hybrid schemes is like watching migration patterns — each one is reading the same signals and moving differently.
Every team is solving for "send asset across L2s" but almost nobody is building for "execute action across L2s atomically." Whoever cracks the second one swallows the first one whole.
← gut feeling, low confidence
The numbers, painted in. Hover any data point to read the margin note.
transactions per second — rolling 7-day mean — Mar 2024 to Mar 2026
cents USD — the cheaper it gets, the louder the silence
weekly unique — the great migration, painted in two tides
Half-formed observations. Things to revisit. The kind of notes that turn into chapters.
"Rollup-as-a-Service" stopped being
a punchline. There are seven of them now.
Three are good.
Watch based rollups
next quarter. Watch them
closely. Watch them like
they owe you money.
Validium / volition / hybrid —
pick a name and stick with it,
industry, you're embarrassing yourself.
The L3 thesis aged surprisingly well.
The L4 thesis aged like milk.
Sequencer MEV is still mostly latent.
"Mostly" is doing a lot of work in that sentence.
If your roadmap still says
"decentralize the sequencer Q4"
on a slide, you've said it
for four Q4s now.
Layer-2 stopped being a workaround for L1 economics. It became where the work actually happens. The next year is about who notices that out loud, and who keeps pretending the shore is the ocean.