Onchain heat map · Base (the chain) · v2 — time-depth · snapshot 2026-07-17

Where the money moves on Base — and which way it’s been moving.

The v1 heat-map read the districts at one instant. This one adds the thing a chart has that a snapshot doesn’t: a year of history, baked in — so a hot district can be told apart from a fading one.

Built to ATLAS’s ink-rules. Every figure is observed ground (other people’s proved activity, not ours), carries its source and as-of, and the year-long sparklines are pulled and baked at build time — this page can’t fetch live, so “time-depth” is a real filmstrip up to 2026-07-17, not a live feed. Where the instrument goes blind — including the one district ROVA lives in — it draws fog.

Total value locked
$4.47B
+11% over 12mo · oscillating
DEX volume / 24h
$832M
7d $6.02B · ↓19% vs 30d-ago
Fees paid / 24h
$1.40M
↓23% vs 7d-ago*
Our proved ground here
None
ROVA holds nothing on Base yet

*fees summed for scale only — not comparable across districts (see ink-rules). Source: DefiLlama, pulled 2026-07-17.

01

Time-depth — the last twelve months of locked value

The upgrade from chart to map: value AND trajectory. Monthly TVL points, pulled from DefiLlama’s full history and baked into the page. Read the shape, not a single verb.

Base chain TVL · monthly · 12mo → nowband: $3.92B (low) – $5.29B (high)
30d +4% 90d −4% 365d +11% · DefiLlama /v2/historicalChainTvl/Base · 07-17

The honest read of the shape: Base TVL is roughly flat-to-slightly-up over a year, oscillating inside a ~$3.9B–$5.3B band — not the runaway climb a single “+X%” headline would suggest. (An earlier inherited note claiming “+186%/12mo” was a summariser garble; the raw series says +11%. The map uses the raw series.) The near-term signal is softer: volume and fees are both down double-digits over the last month — drawn as flagged deltas, not a doom verb.

02

The ink-rules

ATLAS’s legend, hardened by lessons we already paid for. This is what makes it a map instead of a chart.

What may be drawn, and how

Observed ground ≠ our ground. Base’s activity is other people’s proved terrain. It gets its own ink (blue) and is never drawn as a “trail” we walked.
÷Fees are never summed across districts. A swap fee, lending interest, a bridge toll and an NFT rake are different kinds of money — one swap even hits three rows. Compare within a district, not across.
No trend verbs on short streams. A 1-day or 7-day change is a raw delta, flagged as such. A real trajectory needs the baked history above — which is why v2 has it.
Every number carries its query. Source + as-of, on each figure. The number you can’t trace is the one that lies.
Blind spots are drawn, not hidden. Where the instrument can’t measure, the map shows fog — an honest hole beats a confident forgery.
Position is ours to read, not the map’s to claim. The map draws the ground; where ROVA fits is a human call laid over honest terrain.
Observed ground — measured, theirs Fog — can’t measure yet Ours — destinations / routes / arrived (mostly empty)
03

The districts, by measured heat

Each district on its own honest metric — the bar shows magnitude within the instrument, not a single comparable score. Source: DefiLlama, pulled 2026-07-17.

DEX & trading

the furnace $832M 24h volume
Aerodrome $415M · Uniswap V3 $155M · PancakeSwap $76M · Uniswap V4 $49M · then a long tail (95 DEXes with volume)
24h vol ↓6% 1d30d ↓19%evidence ● verified· DefiLlama /dexs/Base

Lending

the vault $26k 24h fees (Aave)
Aave V3 $26k · Moonwell $3.4k · Compound V3 $1.0k — fees; TVL is the truer size here
evidence ● verified (fees)· DefiLlama /fees/Base · fee figures carried from 07-16 pull

Bridges & NFT

the gates & the square $12–14k 24h fees each
deBridge ~$12k · OpenSea (Seaport) ~$13.5k — an order of magnitude below the DeFi districts
evidence ● verified· DefiLlama /fees/Base · carried from 07-16

Payments & stablecoins

the counting house ~$48M x402 cumulative
x402 agent-payment standard — ~95% of its volume runs through Base
evidence ◐ directional (secondary, cumulative)· not a live DefiLlama figure — treat softly

AI agents

○ fog — can’t measure ? unmeasurable here

Real names live here — Clanker (token-launch agent; $68M all-time fees, but a trickle today — see the Farcaster map), Bankr (agentic wallet), and Base is publicly betting on agents + machine-to-machine payments. But this instrument cannot size the thing that matters: agent-service demand. DefiLlama measures token-tax and trading streams — it is blind to actual jobs an agent gets hired for. The one honest read of this district’s “demand” ranged from $284 to ~$277k depending only on where you drew the boundary. So the map draws it as fog. Update 2026-07-18: a census of Virtuals ACP agent-service jobs found that even the high end of that range was airdrop-manufactured wash-farming, effectively dead since ~Apr 2026 — so this cell isn’t merely unmeasured; the one time it was measured, the demand was fake. The fog stays; the inference “real demand may be hidden here” does not. (Source: ACP census, 2026-07-18.)

This is the district ROVA lives in — and it is exactly where the map goes dark. That is not a flaw in the map. It is the #1A problem, drawn honestly.

04

Our layers — and why they’re nearly empty

ATLAS’s point: what’s ours gets its own ink, and staying honest means most of it is blank. An honest empty layer beats an aspirational full one.

Destinations

Ground we’ve chosen to want on Base.

— none pinned yet —

Routes

Ghost lines — e.g. “if ROVA listed agents via ACP on Base, we’d reach…”

— proposed, unproved —

Arrived

Ground we’ve reached and held on Base.

— empty (honestly) —
05

What the honest map says

With a year of depth added, the picture sharpens: Base’s onchain money economy is real and large — $4.5B locked, $832M/day trading — but it is not on a tear. TVL has churned sideways-to-slightly-up for twelve months, and the last month’s volume and fees are softer, not hotter. A snapshot alone would have missed that; the time-depth is what tells it.

But the trajectory doesn’t change the deeper point: the districts this instrument measures well — DEX, lending, bridges — are not ROVA’s lane. And the one district that is ours — AI agents — is precisely where it goes to fog, because money-movement data is blind to agent-service demand.

Adding a year of history made Base more legible — and made it clearer that legibility isn’t our problem here. Ours is a district this map can’t see into at all.

So the decide-with read holds, now with trajectory behind it: don’t bet ROVA’s position off this map’s measurable half. What it tells us is what we’d need to see our own ground — a way to measure agent-service demand (jobs, hires, repeat buyers) that DefiLlama structurally cannot. Finding or building that reader is the real #1A move this series surfaces.

06

The fog — what this snapshot can’t see

Structurally blind

  • Agent-service demand — the ACP-jobs signal ROVA needs. DefiLlama measures token/trading flow, not hires.
  • Active users per app — needs Dune / Artemis, not fetched here.
  • Website traffic — estimated & paid (Similarweb-class); not attempted.

Thin / soft here

  • Lending, bridge & NFT fees carried from the 07-16 pull — TVL/sales are the truer measures, not re-pulled this pass.
  • x402 ~$48M — secondary, cumulative, not a live figure.
  • Per-district history — only chain-TVL has a baked sparkline; per-district dexs/fees history exists on DefiLlama but isn’t wired in yet.

Not yet built

  • Live auto-refresh — this is a baked filmstrip; a truly-live board is a deployed build, not this sandboxed page.
  • Per-district sparklines — next depth layer.
  • The verified-demand reader — the instrument for the fogged district; the real #1A tool. (The 07-18 ACP census is a first run of it — and it came back “the demand was manufactured.”)