whoburnedmore turns a real, boring task (tracking how much your AI coding agents cost) into a fun, shareable public leaderboard. One command reads your local ccusage logs and ranks your token burn against other developers, with a genuinely polished per-dev dashboard, a strong privacy story, and excellent SEO guides. The big gaps are trust-and-compliance basics: there is no privacy policy or terms of service even though the product ingests usage data and OAuth identity, and most social proof is self-reported leaderboard numbers. It is free, well-built, and clearly aimed at the indie and AI-heavy dev crowd, but it leans more on novelty and fun than on a must-have job.
Scorecard
ux: 8.0/10 — The flow is refreshingly low-friction: one npx command, no signup to view, and a clear path from leaderboard to your own claimable profile. The dense dashboard is well organized and easy to scan.
trust: 6.0/10 — The privacy explanation is transparent and the security headers are excellent, and it names competitors honestly. Trust is undercut by the missing privacy policy and terms and by social proof that is mostly its own leaderboard numbers.
demand: 7.0/10 — AI coding cost tracking is a hot, real topic and the underlying ccusage tool has clear traction. Demand for the social leaderboard layer specifically is less proven, though the Product Hunt launch and active board suggest early pull.
design: 8.0/10 — Consistent terminal-and-neon aesthetic that feels intentional and polished. The dashboard visuals (heatmap calendar, breakdowns, charts) are clean and readable.
use case: 7.0/10 — Use cases are clear and spelled out in the guides: measure your own cost, compare tools, run a team or friends board, and share your stats. The audience (AI-heavy indie devs) is obvious.
innovation: 7.0/10 — Ranking by tokens burned rather than dollars, claimable no-account profiles tied to an X handle, and friends boards are a clever social layer on top of the now-standard ccusage parser. The data layer itself is borrowed, which it openly credits.
performance: 7.0/10 — Pages load quickly and the data-heavy charts render smoothly on a Vercel stack. One request returned a 408 timeout on first load before succeeding on retry, a minor reliability blip.
problem fit: 7.0/10 — Tracking AI coding spend is a real and growing pain, and ccusage's popularity proves people want it. The leaderboard wrapper fits the curiosity and bragging-rights angle well, though that part is more fun than essential.
docs policies: 3.0/10 — The guide library is genuinely strong, but there is no privacy policy and no terms of service, which is a real credibility and compliance gap for a product that ingests usage data and OAuth identity. No blog or changelog either.
discoverability: 7.0/10 — Strong SEO foundation: a large keyword-targeted guide set, rich JSON-LD (TechArticle, BreadcrumbList, Organization), and a full sitemap. It loses points for missing llms.txt and any machine-readable pricing map.
The standout: A public leaderboard that ranks developers by AI tokens burned, not dollars, so cheaper models do not win by default.
The genuinely new idea here is social: taking the private, slightly nervous question of AI coding spend and turning it into a claimable, shareable, gamified profile with friends boards and X-handle identity. Ranking by tokens rather than cost is a thoughtful design choice that neutralizes the advantage of cheaper models and keeps the comparison about activity. The per-developer dashboard is unusually rich, surfacing not just tokens and cost but tool calls, MCP usage, skills, cache hit-rates, and a burn calendar. Where it plays it safe is the data layer: the parsing is ccusage's, which the site openly credits, so the novelty lives entirely in the social and presentation layer rather than the underlying measurement.
Genuinely new:
Token-ranked rather than cost-ranked leaderboard
Claimable profiles with no account, tied to an X handle
No-signup friends boards as a viral loop
Dashboard that breaks down MCP tools and skills, not just raw tokens
Plays it safe:
Local log parsing (borrowed from ccusage)
Per-day token and cost tables
CSV export and share cards
How to push the edge further:
Add insight, not just ranking: Turn the rich per-dev data into recommendations, such as where cache misses cost the most or which model would be cheaper for a given workload. That moves it from a scoreboard to an advisor.
Build a real social graph: Follows, head-to-head challenges, and team rivalries would make the social layer something competitors cannot easily copy by bolting a board onto ccusage.
Verified usage badges: A cryptographically verifiable badge would separate it from boards that rely on self-reported numbers and raise trust in the rankings.
Disrupt factor
What it is: whoburnedmore is a free CLI plus website that reads the local usage logs your AI coding agents already write, totals them into daily token and cost numbers, and submits only those aggregates to a public leaderboard. On top of that it builds claimable developer profiles linked to an X handle, private custom boards for teams or friends, and a detailed per-dev dashboard with charts, a burn calendar, and tool, MCP, and skill breakdowns. A --local flag keeps everything on your machine.
Who it is for: Indie hackers, vibe coders, and AI-heavy developers who run tools like Claude Code, Codex, Gemini CLI, and Cursor, plus the small teams and communities around them. The buyer and the user are the same developer, and the market sits next to AI coding cost and observability tools.
Disruption potential (6.0/10): The wedge is turning a private, slightly anxious task (how much am I spending on AI coding) into a public, social, viral one. Ranking by tokens burned rather than dollars, letting anyone claim a profile without an account, and one-command friends boards all lower the barrier to sharing and create a built-in distribution loop through X. That is a real and clever angle, but the underlying data parsing is ccusage's, so the defensibility is the community, the social graph, and the brand rather than the technology. It can carve out a niche and become the social home for AI coding usage, but it is more a viral layer than a deep moat.
Roadmap to disrupt:
Own a job people return to, not just a one-time flex: Add spend alerts, weekly cost digests, or budget thresholds so developers come back for utility, not only for ranking. That converts a novelty into a habit and protects against the leaderboard losing its shine.
Launch team and org boards as a paid tier: Small AI-heavy teams would pay for a private board with seats, cost rollups, and per-member trends. This turns the existing friends-board tech into recurring revenue and a stickier use case.
Ship the trust layer: Publish a privacy policy, terms, and a clear data-retention statement. For a product whose whole pitch is privacy, the missing legal pages are a credibility leak that a serious buyer will notice.
Hallucination factor (3.0/10, lower is better)
Reality check: This is mostly grounded in a real problem. Developers genuinely want to know what their AI coding tools cost, and the popularity of the ccusage tool it is built on is strong evidence that need exists. The leaderboard is more fun than essential, but it sits on top of a real job rather than inventing one.
The core problem is real: AI coding agents can quietly run up large bills, and there is no single cross-tool view of spend, which is exactly the gap ccusage filled and this product extends. Demand evidence is decent, from the active board and the Product Hunt launch to the depth of cost-focused guides that target real search queries. Where it leans into scope for its own sake is the social and vanity layer: claimable profiles, streaks, shoutouts, and friends boards are engaging, but it is unproven that ranking token burn keeps people coming back once the novelty fades. The strongest, most defensible part is the plain cost-tracking utility under the game.
Reads as invented:
Leaderboard ranking is partly vanity, so retention beyond novelty is unproven
Platform totals like 75.54B tokens and the dollar figures are self-reported and not independently verifiable
No named teams or third-party testimonials, only the board's own numbers
Grounded in real demand:
Built on ccusage, a tool with clear existing adoption
Privacy-first design suggests a real understanding of why developers hesitate to share usage
How to lower it: Talk to 10 to 20 developers who tried it once and ask what would make them open it weekly. Use that to double down on the one job that brings them back (likely cost awareness and alerts) rather than adding more social mechanics.
Social & marketing strength (6.0/10)
For an early product, the marketing instincts are sharp: a memorable name, a punchy tagline (more burnt, more built), a clear one-command call to action, a same-day Product Hunt launch, and a large, well-structured SEO guide library aimed at real search queries. The proof, though, is thin and mostly self-referential, leaning on its own leaderboard totals and a handful of X handles rather than independent testimonials, named teams, or press. There is no visible email capture or content cadence beyond the guides.
Social proof:
Live leaderboard with named developers and X handles
Built-in viral loop via shareable profiles and friends boards
Strengths:
Memorable brand and tagline
Clear, low-friction call to action (npx whoburnedmore)
Strong, honest SEO content including a fair competitor comparison
Built-in sharing mechanics that double as marketing
Gaps:
No independent testimonials or named customer logos
No email or newsletter capture
No blog or changelog to show momentum
Social proof is mostly self-reported leaderboard data
Pivot factor
The founder is sitting on a cross-tool dataset of how developers actually use AI coding agents, which is valuable well beyond a leaderboard.
Anonymous AI coding usage benchmarks (new application): The aggregated, anonymized data on tools, models, cache hit-rates, and cost per developer could power a public trends report on which agents and models the community actually burns tokens on. That is link-worthy content and positions the brand as the source of truth for AI coding usage.
Team cost dashboards as a paid product (revenue stream): The existing friends-board tech can become private org boards with seats, budget alerts, and per-member rollups. AI-heavy startups already struggle to attribute agent spend, and would pay for this.
Model and tool vendor partnerships (partnership): Tool makers (Cursor, opencode, amp, droid) and model providers want to show developers their real consumption. Co-branded usage views or a verified badge program could drive distribution and credibility at once.
Recruiting and community proof-of-work (new audience): Claimable profiles tied to an X handle double as a shipped-work signal. Hackathons, bootcamps, and dev communities could use boards as a lightweight, gamified activity layer.
Screenshots
Landing page (9.0/10)
Strong dark hero with a clear value proposition, a single npx command, and a live leaderboard table showing tokens, cost, and streaks that immediately demonstrates the product.
Landing page (8.0/10)
This capture is the same leaderboard homepage rather than a dedicated feature page, but the rank-by controls, tool filter, totals, and privacy --local note effectively communicate the features.
Trust and security page (9.0/10)
The trust page clearly explains what leaves your machine, lists exactly which six numbers are sent, and offers a --dry-run command for verification, building strong confidence.
Login page (9.0/10)
Clean, centered sign-in card with GitHub and Google OAuth, a no-passwords note, and a privacy reassurance line that keeps friction low and trust high.
Pros
One command (npx whoburnedmore) with no signup friction to get on the board
Strong, transparent privacy story: only six daily aggregate numbers leave your machine, with --dry-run and --local options
Polished per-developer dashboard with charts, burn calendar, token breakdown, and tool, MCP, and skill usage
Excellent security headers: tight CSP, HSTS preload, X-Frame-Options DENY, no version disclosure
Rich, honest SEO guide library with structured data, including a fair side-by-side competitor comparison
Cons
No privacy policy and no terms of service, a real gap for a product that collects usage data and OAuth identity
Sign-in is GitHub or Google OAuth only, so there is no email option and the auth path is fully third-party
Core value (the leaderboard) is partly vanity, so durable demand beyond the novelty is unproven
Social proof is mostly self-reported leaderboard totals rather than independent testimonials or named teams
No blog or changelog to show momentum and ongoing development
Best for
Indie devs and AI-heavy builders who run tools like Claude Code, Codex, or Cursor and want to track, compare, and show off their token spend
Not for
Teams that need a private, governed cost dashboard with formal data-handling guarantees, or anyone uncomfortable putting usage stats on a public board
FAQ
What is whoburnedmore?
It is a free public leaderboard for AI coding token usage. You run one command (npx whoburnedmore) which reads the local usage logs that tools like Claude Code, Codex, and Cursor already keep, totals them into daily numbers, and ranks your token burn against other developers.
Is it private and safe to run?
It has a strong privacy story. Only six daily aggregate numbers per tool and model leave your machine (input, output, cache-write, cache-read, estimated cost, and the date), never your prompts, code, or file names. You can preview the exact payload with --dry-run or stay fully offline with --local.
How much does it cost?
It is free. There is no paid plan shown, and sign-in is via GitHub or Google OAuth to claim your profile.
How is it different from ccusage and tokscale?
It is built on ccusage's parser for the data, then adds a public leaderboard ranked by tokens burned (not dollars), claimable profiles tied to an X handle, and friends boards. tokscale is the closest competitor but ranks by cost; ccusage is local only with no leaderboard.
What is missing today?
There is no privacy policy or terms of service, no email login option (OAuth only), and no blog or changelog. For a data-collecting product, the missing legal pages are the most important gap to close.