💛 Support us

Keep The Aporia independent

Everything The Aporia gives you is free. Keeping it free is not. The Aporia exists to make it easier to understand how the news is being covered, without putting that understanding behind a paywall, filling the site with advertising, or asking you to trust whoever happens to fund it. That is why the project asks for donations.

Donate
CA$0 of CA$500 this month
Public information is free.
Computational infrastructure is not.

Reading this site costs nothing and will keep costing nothing. What costs money is the machinery underneath it: the servers, the storage, and the model time that turns raw coverage into something you can compare. A donation pays for that while the side of the project meant to fund it in the long run is built.

What your donation supports

Free today
Everything a reader uses
Nothing here is behind a subscription or a paywall, there is no advertising, and personal data about readers is never sold. What happens if donations do not cover the bill is answered plainly further down this page.

You do not pay for access. You do not pay for more features. You do not pay to see a different analysis.

Programmatic and bulk access
A second side of The Aporia, still being built: what an organisation needs and a reader does not. Intended for universities, researchers, newsrooms, think tanks, NGOs, analysts and businesses.

Paying for access is not paying for influence. An organisation can pay to use The Aporia's data. It cannot pay to change what The Aporia says, how the analysis works, or what conclusions it reaches. That independence is one of the standing commitments on the about page.

API and data

Where the money goes

The goal is to keep The Aporia free without making it dependent on whoever has the deepest pockets.

The organisation-facing side of the project is not generating revenue yet. Until it does, donations are what bridge the gap between the free service and the cost of running it. That is the whole of the ask on this page.

October 2026 0% of goal
CA$0of CA$500
CA$500 still to raise this month, from 0 donations so far.
2026 so far 0% of goal
CA$0of CA$2,000
The marker shows how much of the year has passed (76%). 0% of the annual goal has been raised so far.

The goal is not to build a surplus. It is to cover the cost of running The Aporia and leave enough room to keep improving it.

Why this goal

At its current size, The Aporia's basic operating costs are relatively small. The first call on the goal is what the project cannot operate without. That is the development tools it is built and maintained with, including the Claude Code subscription, the hosting that keeps the website and its services online, and the domain that keeps The Aporia reachable at its permanent address. After those come the tools the project needs as it grows, including Microsoft services and other business infrastructure. Anything beyond the essentials goes toward expanding the project. The long-term intention is to hire students as software engineers, and eventually to run hackathons that introduce more people to The Aporia while helping find people who can contribute to it. So the monthly target is not a salary, a profit target, or a large corporate budget. It is the amount the project currently believes it needs to keep moving.

Help The Aporia analyse more of the news

The Aporia's analysis PC 0% funded
CA$0of CA$3,000
A dedicated PC with a 24 GB graphics card to run The Aporia's language models full time. Today a single home computer analyses about 170 articles an hour and new ones arrive faster than that, which is what stops us adding sources. This machine would analyse far more articles, from more and new outlets, with a larger and more accurate model.
A donation is a voluntary gift, not a purchase.

Card payments are handled by Stripe. The Aporia does not receive or store card details. Full donation terms.

Where donations would go Example

These figures are an example, not a decision. The Aporia has not taken any donations yet, and how the money should be divided is a question that needs money to answer. The split below is a plausible one, shown so you can see the shape of what gets published — not a commitment to these categories or these percentages. The real allocation will be set once donations are actually arriving, and it will be published here in the same form.

The allocation is published so you do not have to take the project's word for it, and publishing an example first is part of that: it shows the format the real one will arrive in, and it can be checked against what actually gets published later. The monthly record starts on 1 October 2026 — from then on, every month's donation total and its breakdown are published here, including the months that raise nothing.

No money is being divided here. These are illustrative shares, shown in the format the real breakdown will use once donations are arriving and the split has been decided.
Servers and hosting
The server the site runs on, its bandwidth, and the backups of the article database.
40%
Model compute
The GPU time behind event grouping, article summaries and leaning scores. This is the largest variable cost, because it grows with the volume of news covered.
30%
Storage and archive
Keeping the full analysed history rather than a rolling window, so current coverage can be compared with how the same story was told months ago.
18%
Development
New sources, new signals, and the maintenance that keeps existing feeds from quietly failing.
12%

The categories and percentages above are an example and have not been committed to. No monthly breakdown is published yet: the first one covers the first complete month after 1 October 2026, and every month after it is published whether or not it raised anything.

These shares apply to donation income only. Anything raised above the monthly goal carries over into the following month.

Where it went, month by month

Donations should not disappear into a black box. Every month, the donation total and the allocation of that month's income are published, and a month with no donations stays on the record rather than being left out of it.

Donations received per calendar month, in CAD. Months with no donations are shown as empty rather than left out. The line marks the monthly goal of CA$500.

Totals cover calendar periods in UTC. October 2026 covers that month only rather than the past 30 days, and 2026 covers the calendar year. Refunded donations are excluded from both figures.

Donations are never published with names attached, and no donor is listed on this page.

The point of all this

Understanding the news should not depend on who can afford it.

The intended future is one where the public side of The Aporia pays for itself through organisations using its data professionally. That is not where it is yet. Until it is, donations are what allow the work to continue without changing what The Aporia is for.

If you find The Aporia useful, even a small donation helps. Thank you for helping keep it independent.

About The Aporia

The Aporia is an independent analysis layer over the news. It collects reporting from different publishers, groups the articles covering the same event, and examines each one using the same set of measurements.

The purpose is to make it easier to see how different accounts of the same event relate to one another.

43,326articles analysed
26,104events identified
51publishers scored
1,118,797claims extracted
53active feeds

How it works

The Aporia produces no reporting of its own. It starts with the event rather than the article: pieces arrive from 53 feeds as they are published, and the ones describing the same real-world event are brought together and treated as different accounts of one thing rather than as unrelated articles.

Each account is then measured against the same fixed questions:

Claim certainty
how strongly the article presents its claims.
Hedging
where language is qualified, softened, or left uncertain.
Political leaning
where the article falls on the left-to-right scale the system uses.
Framing
how the event is characterised and presented.
Corrections
corrections detected in the publisher's own reporting.

Every figure is produced the same way for every source, so two outlets covering one event can be set against each other on equal terms. The measurements are not a verdict on the article. If two outlets describe an event differently, The Aporia does not choose between them: it is not a fact-checker, it never republishes an article's text, and every score here is an automated judgment rather than a statement of fact.

You do not have to take any of that on trust. Every measurement has a defined methodology, the limitations are documented, every article links out to the publisher that wrote it, and where the system is uncertain it says so rather than rounding the uncertainty away.

The point is not to replace your judgment. It is to give your judgment better material to work with.

  1. Outlet A Outlet B Outlet C

    the same event, reported separately

  2. Grouped by the event described

    by what happened, not by keyword

  3. Measured on each Claim certainty Article leaning Publisher corrections

    the same questions, asked of every source

  4. Read side by side

    you compare, and decide

The last box is the one that matters. The Aporia stops at the comparison and leaves the conclusion to you. The methodology page gives the arithmetic for each measurement above.

Why it exists

Most people now meet the news through a feed, and a feed is ranked to hold attention. The most dependable way to hold someone's attention is to show them more of what they already agree with. The problem that creates is subtler than anyone being lied to: over time the range of what a reader sees narrows, until the disagreement on the other side of it looks smaller and less considered than it actually is.

You would not ask a pharmacist to be your physician, even though a pharmacist knows a great deal about medicine. The knowledge is real and the role is different. A single outlet is in the same position. It can be careful, well sourced and worth reading, and still be one account of an event rather than the account of it.

So the useful unit is the comparison. Several reports of one event, set beside each other with the same measurements applied to each, tell you something none of them says on its own: where they agree, where they diverge, and which parts of the story only some of them thought worth including. That last one is usually the most informative, and it is the hardest to notice while reading a single article.

You don't need another news source. You need a way to understand them.

Reading one source

One outlet
▼
One account, with nothing to weigh it against

Reading them together

Outlet
Outlet
Outlet
▼
Where they agree, where they differ, and what only some ran
The right-hand arrangement carries information the left-hand one cannot, however good the single account is, and that holds whichever outlet the single account comes from. Every page here is built to produce the right-hand one.

Who this is for

The Aporia is for people who occasionally want to go deeper than the news usually allows. You might recognise yourself here if:

It is not meant to replace the way you normally keep up with the news, and it is worse than a normal news app at the two-minute catch-up. There is no feed, nothing is ordered by what will hold your attention, and the site will not tell you which account is the correct one. Come here when something catches your attention and you decide it is worth investigating properly.

What this site will and will not do

The Aporia will never
Absolute, because each of these is a decision rather than an effort.
  • Sell personal data about the people who use it.
  • Carry political advertising.
  • Try to move a reader's political beliefs.
  • Hide a material change to the methodology.
  • Let a paying organisation alter a public conclusion.
  • Claim certainty where meaningful uncertainty exists.
The Aporia will do its best to
Worded as effort, because these depend on judgment and time rather than on a single decision.
  • Publish its methodology and keep the documentation public.
  • Publish its financial structure and its major sources of funding.
  • Declare conflicts of interest.
  • Publish significant errors once they are found.
  • Represent uncertainty rather than round it away.
  • Prioritise accuracy over engagement.

The two lists are worded differently because they are two different kinds of promise. The left column is absolute because nothing on it requires effort to keep, only a decision not to do it. The right column asks for judgment, and a project that promised those without qualification would be overclaiming on the same page where it asks you not to take claims on trust.

The Aporia is free to read and the intention is to keep it that way. It is paid for by donations, and the intended future is one where it largely pays for itself through organisations using its data professionally.

If those two do not cover what it costs to run, subscriptions or advertising may be added rather than the project quietly closing. That is not the plan and it is not wanted, but it is the honest answer to what happens if the money does not arrive, and saying so in advance is better than a reader finding out by meeting it.

If it happens, it will be said here first and the commitments above hold regardless: no political advertising, no selling of personal data about readers, and nothing anyone pays for will change a public conclusion or a published score. What the project costs to run, and what any donation goes toward, are itemised on the support page.

Our mistakes

This project publishes what it gets wrong, in a record anyone can read: what a reader would have seen, when it started, when it was noticed, and how long it took to fix. The last two are the ones that say whether a project is careful, so an entry cannot look tidy by leaving them out.

The record is currently empty, and that is not a claim that no mistakes have been made. This site treats an empty correction record as an open question rather than a clean one, because a record can be empty when nobody is checking, and the same reading applies here. Read the record.

Start here

💬Give feedback
🕘History 🎫Support