A Retreat Executed as a Pivot

Brave's fourth BAT roadmap moves the token from paying for attention to paying for things — and sunsets the Brave Ads notification unit, the use case BAT was sold to power in 2017. What is new is a buyback promise whose percentage is never stated, routed into a pool Brave's own terms call company-owned and discretionary, disclosed on a page that has listed no BAT purchase in 278 days. The retreat did not start this July; it started in January 2023.

The Numbers That Frame It all from Brave's own pages or two market sources

278days since the last BAT purchase Brave listed
−81.4%BAT purchased, 2025 vs 2024 (rows shown)
−96%BAT price vs its Nov 2021 high
120.9Mmonthly users — the browser is working

The Two Readings why "pivot or retreat" is a false binary

What is ending
The Founding Use Case
launched 2019, gated from 2023
- Earning BAT required KYC from Jan 2023; virtual BAT was killed and unconnectable users lost balances.
- Users reported per-ad rewards falling 0.025 → 0.003 BAT (−88%).
- The published user share became "greater than or equal to Brave's share" — as little as 50%.
What replaces it
An Unquantified Promise
buybacks, but no number
- "A percentage" of net revenue — the figure is never given, in a post full of other figures.
- Destination is the company-owned User Growth Pool, not a burn address.
- Language is "will fund", "plans to"; the terms carry no buyback obligation.

Bottom line

Brave the company is succeeding; BAT the token is being repositioned for the fourth time. The replacement asks holders to trust an unquantified share of undisclosed revenue flowing into a discretionary pool — a claim about the future backed by less disclosure than the thing it replaces.

起承 · Context · Content 1 / 2

起 — What Roadmap 4.0 Proposes seven initiatives, most of them deferred

Initiative	Substance	Status
Rewards Card	Virtual + physical card on GENIUS-compliant stablecoins, BAT cashback.	issuer, geography, rates all deferred
BravePay	Stablecoin payment layer, `.brave` addresses, "private transactions".	merchants + SDK "longer term"
Creator protocol	Micro-royalties when AI products access registered creators' content.	a blog post promised later in 2026
x402 + MPP	HTTP 402 agentic payments; "equal first-class support" for two rival stacks.	experimental repo
Buybacks	"A percentage of Brave's net revenue" funds BAT buybacks via the UGP Reserve.	tokenomics post deferred

承 — The Retreat Is Three and a Half Years Old the sunset is the last step, not the first

24 Apr 2019Brave Ads with user rewards launches — BAT's founding use case goes live. 7.7 years from here to the sunset.
13 Jan 2023A KYC'd custodial account becomes mandatory to earn. Users who cannot connect get a blanked-out earnings counter.
~Apr 2023Virtual BAT sunset. Users in countries with no provider — Brave's own example is Germany — lost remaining balances.
9 Mar 2023Forum thread reports an 88% cut in per-ad rewards. Brave's VP confirms rewards float with advertiser bids and the BAT price.
21 Oct 2025The last BAT purchase Brave has listed anywhere. 159,174 BAT. Nothing since.
9 Jul 2026The notification ad unit is formally sunset "by the end of 2026".

The purchase record is Brave's own, and it is the strongest single data point. Reading only the rows visibly confirmed on the transparency page: 11 purchases totalling 4,606,715 BAT in 2024, 6 totalling 858,096 in 2025, and none in 2026. The last is dated 21 October 2025 — 278 days before this brief, and 252 days before the page's own "as of Jun 30, 2026" stamp, so the page is being maintained while the table is not moving. Brave is promising new revenue-funded buybacks at the moment its existing, advertiser-funded purchase programme appears to have stopped. And a structural absence matters as much: Brave has never published BAT paid to users, BAT paid to creators, unclaimed BAT, or the UGP balance — so nobody outside the company can compute whether the attention model worked.

52% → ?rewards wallets as a share of users, 2021 vs now (undisclosed)
1.81Mcreators counted on Brave's own page, vs "over 2 million" claimed

轉合 · Pivot & Resolution · Content 2 / 2

轉 — The Destination Is Weaker three checks vs the documents

Claim	As presented	What the documents show
Buybacks	Revenue-funded BAT buybacks supporting rewards and growth.	No percentage stated; "net revenue" undefined for a private company; destination is a pool the terms call company-owned and usable "in its discretion" — a balance-sheet transfer, not a burn. No obligation in the terms.
Agentic payments	x402 and MPP support; a growing space.	Four months earlier the flagship rail cleared ~$28,000/day (~$0.20 per transaction); an analyst called it "mostly a mirage" with ~50% gamed and target merchants "rare at this stage".
Privacy	Preserving "the anonymity … of cash in your back pocket"; "private transactions".	GENIUS-compliant stablecoins must by statute be seizable, freezable and burnable, and a card needs a compliant KYC issuer. The post does not address the tension.

The Creator Protocol Is a Late Entrant same HTTP 402 mechanism, wrong layer

Scheme	Since	Where it sits
Cloudflare pay-per-crawl	2025, expanded Jul 2026	The CDN edge, across roughly a fifth of the web. HTTP 402; publishers set a price, unpaid bots are blocked.
RSL	Sep 2025	Server-side licence terms, launched with Reddit, Yahoo, Medium, Quora, Ziff Davis and The MIT Press — already supporting pay-per-crawl and pay-per-inference.
Brave Creator Contribution	announced Jul 2026	The browser — but AI training crawlers hit servers and CDNs, not browsers, and the post does not explain how a browser-side protocol bills them. Unspecified and unbuilt. Brave's real edge is its 1.81M registered creators.

合 — What to Watch falsifiable tests, by value

Watch for one disclosed buyback — a date and an amount on the transparency page is the cheapest test that 4.0's central promise is real.
Watch whether the deferred tokenomics post names a percentage. Without a number the promise stays unfalsifiable.
Watch whether bought-back BAT is retired or re-spent. A burn address would change the analysis; a company pool would not.
Watch for a card issuer and a geography — everything commercially load-bearing is currently deferred.

Two things are true at once, and they are frequently conflated. The company side is strong: 120.9M monthly and 51.2M daily users — a 42.3% ratio most consumer software would envy — an independent search index at roughly 1.6bn queries a month, and revenue estimated to have doubled to about $52M in 2024. (Revenue, funding and headcount are third-party estimates; Brave discloses none.) Roadmap 4.0 is not a rescue. The token side is nine years old and roughly back where it started: about $0.075, a ~$112M market cap, 96% below the November 2021 high, and within roughly 4–14% of its 2017 all-time low. Roadmap 4.0 also silently drops "The On-Chain Era" — Roadmap 3.0's headline nineteen months earlier — mentioning neither Boomerang, nor Themis, nor disintermediation. Items here are not cancelled; they stop being mentioned, which is the most useful thing to know when reading a fourth roadmap. One caveat on tone: no substantive published criticism of 4.0 was found, and post-announcement community forums could not be reached — so that is "not found", not "did not happen". The critique above is this brief's reading of primary documents, not a report of a backlash.

$1.92 → $0.075BAT, Nov 2021 high to 26 Jul 2026
4throadmap in nine years; 3.7y, 3.8y then 1.6y apart