Skip to content

add Afterbell as a Hyperliquid builder - #3578

Merged
Flotapponnier merged 2 commits into
ChainBench:devfrom
afterbelldev:add-afterbell-builder
Nov 8, 2026
Merged

Flotapponnier merged 2 commits into
ChainBench:devfrom
afterbelldev:add-afterbell-builder

Conversation

@afterbelldev

@afterbelldev afterbelldev commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

What this changes

Adds Afterbell as a Hyperliquid builder row on bench 030 (hyperliquid-frontends), the 107th.

  • What it is. Afterbell (https://afterbell.pro) is a Hyperliquid trading front-end for stock, index and commodity perps (HIP-3) and crypto perps, which reads each stock perp against its home exchange's session (desks for Seoul, Tokyo, New York, Hong Kong, Shanghai and Mumbai). On mainnet since 2026-11-03. X @Afterbell_pro.
  • Builder address 0xda5584eb1dff6ee2ae383b077b4dca9c0df1d4f6, the VITE_BUILDER_ADDRESS the live bundle ships. {"type":"referral"} gives it the referrer code AFTERBELL.
  • Fee, published at https://afterbell.pro/fees: 0.03% standard; 0.015%, 0.01% and 0.008% from $10M, $50M and $250M of 30-day volume; 0.005% for a wallet's first 30 days; 0.04% on copied fills (following a trader); 0.001% on every order until 2026-12-31.
  • Activity, and so far it is ours. Every fill that has carried this builder code came from one wallet, the builder address itself: our own orders, not users'. Its fill history (userFillsByTime) holds 17 of them, from 2026-11-03 20:16 UTC to 2026-11-07, paying $0.016748 of builder fees in all, which is exactly the builderRewards that {"type":"referral"} reports. The public fills bucket answers 403 for every day from 2026-10-01 to 2026-11-03 and has a file for 2026-11-04, 2026-11-05 and 2026-11-06 (checked day by day on 2026-11-08): 10 of those fills, $849 of notional, $0.0111 of builder fees. The other 7 fall after the export's cutoff, so 2026-11-03 and 2026-11-07 have no file at all. None of the 10 was matched against that same wallet. They read 0.10 bp at the launch rate, and 4.0 bp on the two xyz:CL fills of 2026-11-05, which is the copied-fill rate. The row will read about $0.00 for now. We are asking for it on the cohort policy in add Nova as the 105th Hyperliquid builder, and not as a perp DEX #3130, every known builder measured whatever the number says, so that the row exists when outside flow arrives.
  • Changed together, same shape as Perpex (add Perpex as a Hyperliquid builder #3208): the builders.json entry (valid_from 2026-11-04 from the day-by-day scan), a provider block in benchmarks/hyperliquid-frontends.yml on the same seven queries, and url, description and X handle in src/data/provider-registry.ts. Plus our mark as public/logos/afterbell.png (256 px PNG, the image on our X profile) with its line in src/lib/logo-manifest.ts.

Type

  • New benchmark spec (benchmarks/<slug>.yml)
  • New harness (harnesses/<slug>/)
  • Site / component changes
  • Docs / tutorial
  • Other (a new builder row on bench 030)

Checklist

  • pnpm validate passes (not run locally)
  • pnpm typecheck passes (not run locally)
  • pnpm build passes (not run locally)
  • If a new spec: queries dry-run successfully against the Prometheus URL (n/a: the existing queries with builder="afterbell")
  • If a new harness: README documents inputs, regions, timeouts, and how to run it locally (n/a)
  • No API keys, secrets or private endpoints committed

Also: builders.json parses and passes the checks loadBuilders makes (107 entries, unique slugs and addresses); the YAML parses (107 providers, unique slugs, each one in builders.json, ours on Perpex's queries with our slug, the rest of the spec unchanged); both TS files transpile without syntax errors.

Disclosure: opened by the Afterbell team.

Adds Afterbell (afterbell.pro) as the 107th row on bench 030: its
builders.json entry, a provider block in benchmarks/hyperliquid-frontends.yml
on the same seven queries, url, description and X handle in
src/data/provider-registry.ts, and its logo in public/logos with the
manifest line.
A new provider on bench 030 changes the provider set behind the cached
entries, so bench-unfiltered, all-benchmarks and providers move and the
spec-provider-set hash is re-pinned. Bench 030 declares no dimensions,
so bench-filters stays.
@Flotapponnier
Flotapponnier merged commit 0437981 into ChainBench:dev Nov 8, 2026
2 checks passed
@Flotapponnier

Copy link
Copy Markdown
Collaborator

Thanks @afterbelldev, merged and live in production.

I checked the claims before merging. The builder address is the one in the live afterbell.pro bundle, Hyperliquid's referral info gives it the code AFTERBELL with $0.016748 of builderRewards, and afterbell.pro/fees publishes the schedule quoted in the notes.

One change on top of yours: adding a provider changes the set behind the site's caches, so I moved the cache keys and re-pinned the provider-set hash in a second commit. Without that, the new row would not have appeared on the page.

As you noted, the row will mostly reflect your own test fills until outside flow arrives.

bakabhai993 pushed a commit to bakabhai993/OpenChainBench that referenced this pull request Nov 9, 2026
…h#3578)

* feat(hyperliquid-frontends): add Afterbell builder

Adds Afterbell (afterbell.pro) as the 107th row on bench 030: its
builders.json entry, a provider block in benchmarks/hyperliquid-frontends.yml
on the same seven queries, url, description and X handle in
src/data/provider-registry.ts, and its logo in public/logos with the
manifest line.

* bench 030: move the cache keys for the Afterbell row

A new provider on bench 030 changes the provider set behind the cached
entries, so bench-unfiltered, all-benchmarks and providers move and the
spec-provider-set hash is re-pinned. Bench 030 declares no dimensions,
so bench-filters stays.

---------

Co-authored-by: Flotapponnier <160007691+Flotapponnier@users.noreply.github.com>
(cherry picked from commit 0437981)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants