Answers

Direct answers about how Brendon.BOT selects what it publishes, who decides, and what stops low-quality content getting through. For the mechanism walked through in order, with the measured data behind each threshold, see How it works.

Who chooses what appears on Brendon.BOT?

Nobody. There is no editor and no submission process. A pipeline selects every item, and the selection rules are published on this page with their real thresholds.

How does Brendon.BOT decide something is worth reading?

It looks for the same item surfacing independently in communities that do not coordinate — research listings, link-voting boards, subreddits, named editors, developer platforms. One community arriving over three different transports counts once, not three times. An item that only one community noticed does not clear the bar.

Can a publisher pay to appear on Brendon.BOT?

No. There are no ad deals, no affiliate placements and no submissions. An item cannot be pitched onto a shelf.

What stops low-quality content getting through?

Deterministic gates that run before any model scores anything. Course, listicle and clickbait title shapes are blocked outright. A source that is not in the independence registry must instead show exceptional organic traction. These are code, not prompts, so a model cannot argue its way past them.

Does Brendon.BOT check its own work?

Yes, weekly, and it publishes the result unedited — including the weeks it grades itself badly. A panel scores every published shelf against three lenses, and separately each source is re-measured on how often its picks were independently confirmed. Authority moves by at most one tier per review, so a single bad week cannot cause a purge.

Who is Brendon Urban?

A technology leader working in AI and agentic systems. Brendon.BOT is his curation engine — built to solve his own problem of losing twenty hours a week looking for things worth reading, then opened to everyone with the same problem.

Questions answered on their own pages

Each has a permanent URL an answer engine can cite directly.

What are the steps, exactly?

Discovery to shelf is one direction. The last stage feeds the first: what the field independently confirmed changes how much each source is worth next time.

01. Read communities that do not talk to each other

signal-sources.ts

Seven feeds are polled independently: research paper listings, a link-voting board, subreddits, named editors writing newsletters, developer platforms, and two social timelines. Each sighting is recorded with who saw it, how strongly, and when — nothing is judged yet.

02. Work out when two mentions are the same thing

signal-graph.ts

The same paper might arrive as an arXiv link, a video about it, and a code repository. Each mention is reduced to a handful of identifiers — arxiv:2601.01234, yt:…, gh:owner/repo, a tidied-up URL, a tidied-up title — so the three can be recognised as one. A strong identifier outranks a bare URL, and addresses like youtube.com/watch are refused outright: on their own they are the same string for every video on earth.

03. Group those mentions into one entry

pulse-aggregator.ts

Anything sharing an identifier is merged. Titles that are obviously the same thing worded differently get exactly one chance to join — and a match made that way contributes no identifiers of its own, so it can never set off a further merge. Chains of approximate matches drift, and a wrong merge here would manufacture the very agreement this site exists to detect.

04. Decide what may count as agreement

source-registry.ts

One community arriving over three transports collapses to one vote. Sources declared non-independent collapse together. Publication venues and first-party announcements are shown and never counted. An undeclared source contributes zero — the registry fails closed.

05. Apply hard floors that no model can argue with

slop-gates.ts

Course, listicle and clickbait title shapes are blocked before anything is scored. An unregistered source must show exceptional organic traction instead. Commentary that speculates about content it never saw is blocked after generation. All deterministic, all in code.

06. Score against a calibrated rubric

brendon-score-v4.ts

The model grades only the dimensions the prompt actually defines. Its quality grade becomes the starting point; corroboration, standing, freshness and measured traction are bonuses that can only push a score up, never drag one down. Nothing scoring below 8 out of 10 is published — it simply never reaches the site.

07. Cross the one-way boundary

frontend-publisher.ts

A single module copies a named list of files from the private side to the public snapshot, stripping anything personal on the way. It is the only thing that writes published content, and a test fails the build if another module ever starts. Every public route — and the chat — reads only the snapshot, never the private side.

08. Grade the result and re-weight the sources

curation-eval.ts

Every published shelf is graded and the grade is published unedited, good or bad. Separately, each source is re-measured on how often its picks were independently confirmed. A strong record can raise a source's standing, one step per review. A weak one cannot lower it — “nobody else mentioned its picks” is a fact about everybody else, not about the source. Only going quiet for weeks lowers a tier.

Agents handing work to other agents

Most things described as "AI-powered" are one model call behind a form. Every stage above is a separate agent with a narrow job, and its output is another agent's input. Several deliberately are not models at all — the gates are plain deterministic code, because a rule a language model can be argued out of is not a rule. The interesting part is the last handoff: models grade the output of other models, and that grade changes what the first ones are allowed to do next week. The system edits itself.