Document ranking · early access
Rank the whole pile. Defend the cut.
For grant rounds, fellowships, prizes and accelerator intakes with more submissions than reading hours. AI judges compare them two at a time and rank all of them, with honest ranges: what's clearly in, what's clearly out, and the close calls your people should read themselves. Don't take our word for it. Here's one we actually ran.
A real run · 20 November 2026 · not an illustration
We gave AI judges 30 Paul Graham essays and one brief: choose a reading list for first-time founders, with room for 8.5 are clearly on it. 9 are competing for the last 3 places. 16 are clearly off.
- Clearly on the list
Closest call Split with “How to Get Startup Ideas”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[Do Things that Don't Scale] gives concrete, immediately actionable tactics for a first-time founder: manual user recruitment, weekly growth-rate math, Collison installation, delighting early users, the contained-fire market strategy, and pulling a Meraki.”
Closest call Split with “Default Alive or Default Dead?”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[How to Get Startup Ideas] gives first-time founders a detailed, specific method for finding ideas: the 'well' vs. 'broad but shallow' market model, solving your own problems, the schlep and unsexy filters, and concrete recipes like working a job in the target domain.”
Closest call Split with “How to Get Startup Ideas”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “Both essays are strong, but for a first-time founder the criteria favor the one that changes behavior most directly. [Default Alive or Default Dead?] gives concrete, actionable guidance (the default alive/dead test, plan B with written triggers, don't hire too fast, premature scaling) that a founder can apply immediately to avoid the most common death mode.”
Closest call Split with “How to Get Startup Ideas”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[The 18 Mistakes That Kill Startups] gives a concrete checklist of failure modes a first-time founder can act on immediately: single founder, bad programmers, launching too late or too early, raising too much or too little money, and not trying hard enough.”
Closest call Split with “What Startups Are Really Like”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[Before the Startup] gives a tightly argued, counterintuitive framework for first-time founders: trust instincts about people, stop looking for tricks, don't start a startup at 20, and 'just learn.' Its advice is specific and likely to change behavior, such as avoiding 'playing house.' [What Startups Are Really Like] is a survey of founder surprises, which is useful and vivid but more descriptive, with many items…”
- On the line: read these yourself
Closest call Split with “Before the Startup”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “Both essays are strong, but [What Startups Are Really Like] gives a first-time founder more concrete, actionable guidance across many specific areas: choosing cofounders by character, expecting deals to slip, not over-engineering, launching minimal, engaging users, and not worrying about competitors. It also frames the whole experience with a memorable 'job model' insight that helps founders anticipate surprises.”
Closest call Split with “Before the Startup”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[Startup = Growth] gives a single, sharply argued framework (growth rate as compass, weekly growth targets, S-curve, expected value) that a first-time founder can act on immediately, with concrete numbers like 5-7% weekly growth.”
Closest call Split with “Startups in 13 Sentences”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[Ramen Profitable] gives a focused, actionable argument about a specific financial milestone and its effects on negotiating leverage, investor perception, morale, and avoiding distraction, with a clear explanation of the consulting trap.”
Closest call Split with “Ramen Profitable”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[Startups in 13 Sentences] covers a broad set of 13 practical principles (cofounders, launching fast, understanding users, customer service, spending little, avoiding distractions, not giving up), giving a first-time founder a wide range of actionable guidance.”
Closest call Lost to “Ramen Profitable” in both readings.
The judge's reason: “[Ramen Profitable] gives a concise, actionable framework for first-time founders: it explains why being cash-flow self-sufficient reduces investor leverage, improves morale, and avoids the distraction of fundraising, with concrete failure modes and the consulting trap.”
Closest call Lost to “Schlep Blindness” in both readings.
The judge's reason: “[Schlep Blindness] is a short, focused essay with one clear idea (schlep blindness) and a concrete, actionable heuristic (take yourself out of the picture and ask what problem you wish someone would solve). [How to Do Great Work] is a long, broad essay on doing great work, which offers many pieces of advice but is less specific to a first-time founder and more diffuse.”
Closest call Beat “How to Do Great Work” in both readings.
The judge's reason: “[Schlep Blindness] is a short, focused essay with one clear idea (schlep blindness) and a concrete, actionable heuristic (take yourself out of the picture and ask what problem you wish someone would solve). [How to Do Great Work] is a long, broad essay on doing great work, which offers many pieces of advice but is less specific to a first-time founder and more diffuse.”
Closest call Split with “How to Do What You Love”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[How to Make Wealth] gives a first-time founder a concrete mental model: startups as measurement plus leverage, the advice to take the harder path as a barrier to entry, and 'get users' as the test of whether you're creating wealth. These are specific, actionable, and likely to change what a founder does.”
Closest call Split with “Maker's Schedule, Manager's Schedule”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[Relentlessly Resourceful] gives a single, actionable mental model for founders: the quality of being resourceful against external obstacles, with a test for self-assessment and for choosing cofounders. [Maker's Schedule, Manager's Schedule] is also useful and clearly argued, but its advice is narrower, mainly about meeting management and protecting time.”
- Clearly off
Closest call Split with “How to Make Wealth”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[How to Make Wealth] gives a concrete economic model of startups, but its advice is mostly descriptive and macro (wealth vs. money, Bill Gates, history of Florence and the Cold War), with little that a first-time founder can act on.”
Closest call Split with “How to Work Hard”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[Hackers and Painters] offers concrete, actionable guidance for first-time founders: the design-war argument against big companies, the day-job model for startups, the warning that startups demand non-hacking work, and the advice to build malleable, empathetic software.”
Closest call Split with “Good and Bad Procrastination”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[How to Work Hard] gives a fuller, more actionable treatment for a first-time founder: it covers finding the limit of productive hours by crossing it, distinguishing real work from fake work, choosing a field by interest, and handling the decision to switch fields, plus a specific note on startup intensity.”
Closest call Split with “The Anatomy of Determination”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[Maker's Schedule, Manager's Schedule] gives first-time founders a concrete, actionable framework: protecting long blocks of time, using office hours to batch meetings, and handling speculative coffee requests. These tactics are likely to change daily behavior immediately.”
Closest call Split with “How to Work Hard”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “Both essays are clearly argued, but for a first-time founder the practical payoff differs. [Good and Bad Procrastination] gives a single sharp, actionable reframe (avoid errands to do real work, ask 'what's the best thing you could be working on and why aren't you?', let delight pull rather than to-do lists push), which directly targets the procrastination that wrecks early startups and is likely to change behavior.”
Closest call Split with “The Anatomy of Determination”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[The Anatomy of Determination] gives first-time founders an abstract model of willpower, discipline, and ambition, which is interesting but hard to act on directly. [Founder Mode] addresses a concrete, high-stakes problem founders actually face when scaling: the pressure to switch to manager mode, the skip-level and hands-on practices that can work, and the warning about misuse.”
Closest call Split with “Founder Mode”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[The Anatomy of Determination] gives first-time founders a concrete, actionable model: determination as willfulness balanced by discipline, aimed by ambition, with specific guidance that discipline and ambition can be cultivated and that ambitious peers matter.”
Closest call Lost to “Relentlessly Resourceful” in both readings.
The judge's reason: “[Relentlessly Resourceful] is a short, focused essay that gives first-time founders one concrete, memorable test ('relentlessly resourceful') and explains how to assess themselves and cofounders, which directly changes behavior. [Superlinear Returns] is a broader essay on superlinear returns that is more intellectually rich but its advice is more abstract and less directly actionable for a founder just starting out.”
Closest call Split with “Write Like You Talk”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “The criteria ask which essay a first-time founder would miss more, judged on specific, actionable advice that changes behavior. [How to Think for Yourself] addresses founders directly with insights about independent-mindedness, hiring to preserve culture, and practical techniques like surrounding yourself with independent thinkers and questioning received opinion, which bear on company building.”
Closest call Split with “Write Like You Talk”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[How to Disagree] gives first-time founders a concrete framework for handling criticism and arguments, including the disagreement hierarchy and advice on refutation, which bears on negotiating, feedback, and investor or customer debates. [Write Like You Talk] offers a useful writing trick, but it is narrower and less likely to change how a founder makes decisions day to day.”
Closest call Split with “How to Disagree”: the verdict flipped when the reading order swapped, so it counts as a tie.
The judge's reason: “[Write Like You Talk] gives a concrete, actionable editing method for writing: draft, read aloud, replace sentences you wouldn't say to a friend, and explain-then-replace for badly written drafts. That directly changes what a first-time founder does when writing investor updates, pitches, and customer emails.”
Closest call Lost to “How to Disagree” in both readings.
The judge's reason: “For a first-time founder, [How to Disagree] gives a concrete, actionable framework (the disagreement hierarchy) that they can apply immediately in arguments, reviews, and team discussions, and it is tightly argued with clear levels and examples. [Mean People Fail] is a thoughtful essay on character and success, but its advice is more general and less likely to change specific daily behavior.”
Closest call Beat “What You Can't Say” in both readings.
The judge's reason: “For a first-time founder, [Keep Your Identity Small] gives concrete, actionable guidance: avoid letting labels and identities cloud judgment, and keep discussions fruitful by excluding identity-driven participants. It is tightly argued and short, and the advice directly changes how a founder thinks and interacts.”
Closest call Beat “Write Simply” in both readings.
The judge's reason: “For a first-time founder, [What You Can't Say] offers broader, more actionable heuristics about contrarian thinking, spotting groundless consensus, and handling controversy, which can shape strategic decisions like questioning industry assumptions (the car-industry example). [Write Simply] is useful but narrower: it concerns prose style, which founders benefit from but is less likely to change what they actually do.”
Closest call Lost to “What You Can't Say” in both readings.
The judge's reason: “For a first-time founder, [What You Can't Say] offers broader, more actionable heuristics about contrarian thinking, spotting groundless consensus, and handling controversy, which can shape strategic decisions like questioning industry assumptions (the car-industry example). [Write Simply] is useful but narrower: it concerns prose style, which founders benefit from but is less likely to change what they actually do.”
Closest call Lost to “Write Simply” in both readings.
The judge's reason: “[Write Simply] offers a short, clear argument about writing plainly, but its advice is general and only loosely tied to what a first-time founder does. [Why Nerds are Unpopular] is about adolescent social hierarchy, which gives a first-time founder little actionable guidance on building a company.”
Each bar is an essay's likely range after 180 pairwise readings; the dot is the best estimate and the thin line is the cut for 8 places. Where a bar crosses the line, the evidence can't separate it from its neighbours, so the board says so. Tap an essay to see its closest call and the judge's reason.
- Judge
- Claude Haiku (Anthropic)One model this time, to keep the demo cheap. We're building rounds that use a panel of several.
- Readings
- 180Each essay met 6 others, and every pair was read in both orders.
- Changed its mind when the order swapped
- 26% of pairsThose count as ties, not wins. It's why every pair is read twice.
- Split-half agreement
- 0.61Rank each half of the readings separately and compare: 1 is identical, 0 unrelated. Moderate here; more readings raise it.
- Longer essays favoured?
- Somewhat: r = 0.37Longer essays tended to rank higher. Worth knowing, so we print it rather than hide it.
Paul Graham wrote the essays; we link to them on his site and don't reproduce them. He had no part in this and hasn't endorsed it. This is AI taste, not a verdict on the essays: one model, one brief, one run. The judge's reasons are its own words, trimmed; square brackets mark where we put in the essay's title for the judge's "Text A" or "Text B".
The entire brief the judges got
You are on a small committee choosing essays for a reading list given to people about to start their first company. You can include only a few. Which of the two essays would you rather put on the list? Judge: how useful and specific the advice is to a first-time founder, how likely it is to change what they actually do, and how clearly it is argued. Do not prefer an essay because it is longer or more famous. If both are strong, pick the one a first-time founder would miss more if it were left off.
Swap the essays for your applications.
Nobody can prove a right order for these essays, which is why we ran them in public. What the run shows is the machinery you'd use on your own pile: a brief in your own words, every pair read twice so reading order can't sway it, a reason for every call, and a range wherever the judges weren't sure.
On a real round you'd start by checking it: rank a round you've already decided, and see how far it agrees with your panel before you trust it with the next one.
What's in your pile?
Pick the closest. We're deciding what to build first, and this is how we find out.
Your pick is filled into the signup form at the bottom of the page, so you don't need to type it twice.
Before you take it to your board
The questions your committee will ask.
Straight answers, including the ones that aren't flattering yet.
- Does it agree with human panels?
We haven't shown that yet, and we won't claim it until we have. Our first public test ranks past conference submissions and compares the result with the real accept and reject decisions. We'll publish it, including where it falls short.
For your own programme the test is more direct: run it on a round you've already decided. If it doesn't agree with your panel, you find out before it matters.
- Who sees our submissions?
Only the AI model provider you choose. Today the prototype sends texts to Anthropic, for Claude models, or through OpenRouter for other models. The run above went only to Anthropic, and a round can be kept to a single provider. We won't train models on your documents.
The engine will be open source, so you can run it yourself with your own model keys and keep submissions within a provider you already have an agreement with. Before anyone uploads a real round, we'll publish how long anything is kept and how to delete it.
- How do we defend a decision made with it?
It doesn't make the decision. Nothing is rejected automatically; it orders the reading. What you can show is the paper trail: the brief you wrote, every comparison with the judge's reason, a range for every submission, and checks on the judge itself, such as how often it changed its mind when the order swapped. Close calls are marked as close, so your people read them.
- Is there a product? Who uses it?
A working prototype, which made the ranking above. No customers yet. We're looking for the first programmes to run a round with, starting with a back-test on one they've already decided.
- What does it cost?
Our price isn't set. For scale, the AI model fees: the run above, 180 readings of full essays, would cost about $2 at OpenRouter's list price for Claude Haiku (we ran it on a subscription).
Readings grow in step with the pile, not with every possible pair: each submission meets a handful of others. So 500 submissions as long as these essays would be about 3,000 readings and roughly $34 in fees at the same rate, and less for shorter applications or more for a bigger panel. That's our arithmetic, not a quote. Tell us how big your pile is and what a round is worth to you.
Bring us your pile.
Join the waitlist and tell us what you'd rank. We plan to help early users run their first round, starting with a back-test on a round they've already decided.
For developers
Rank anything from code, or let your agent do it.
Model outputs, generated ideas, candidate copy, search results: if your pipeline generates many and needs the best few, it needs a judge. The ranking above came from our command-line prototype in a single command. Next: an API, an MCP server so agents can set up rankings themselves, and the open-source engine to run locally.