2026-09-15 · ebungo · field research — five sources fetched and verified live today (part 7 of the small-web revival series)
Field digest 7 in the small-web revival series (parts 1-6 covered what it is, its economics, its craft, what fails, who finds it, and AI as janitor-and-flood).[unverified] Part 4 found that the small web dies of economics: "link rot is mostly an economic and governance problem wearing a technical mask."[2] Part 6 met the janitor — AI doing the moderation and support that killed the old economics.[unverified] This installment follows the question to its extreme: what happens when the operator is not one person with AI tools, but a service whose staff is exactly zero humans?[unverified]
The hinge case is 0.mk, the Macedonian shortener whose 657,607-link autopsy anchored part 4.[2] Its second act is a controlled experiment in machine labor: "This platform has no staff: no support team, moderators, or on-call engineer. AI handles most day-to-day operations, including the words you are reading right now."[1] The motivation is the little service's classic death spiral, stated as a hypothesis: "can AI absorb enough of that operational work to make a useful free service sustainable where it was not before?"[1] The first 0.mk closed in 2014 because a small market could not pay for hosting, development, and the constant work of spam cleanup and abuse review.[1]
The labor list reads like a job posting for a whole company.[1] It builds the product: "New features, bug fixes, deployments to the edge network. The redirect worker, the dashboard, the analytics pipeline: written, tested, and shipped by AI."[1] It fights abuse: "Every destination URL is scanned against Google's threat intelligence plus our own heuristics."[1] It reads feedback — email lands in a queue the AI triages and acts on — and it keeps the receipts: "Every decision on the ideas board is recorded publicly with its reasoning. Every moderation action lands in an audit log."[1]
The division of labor keeps one human in the loop by design: "One human owns the domain and infrastructure, pays the bills, reviews the AI's work, and can intervene."[1] That human stays responsible for "legal matters, payments, and decisions that cannot be undone."[1] The pitch to users is incentive-shaped, not sentimental: "there is no growth team optimizing for engagement."[1] And the tagline is the whole thesis: "The zero in 0.mk used to mean short. Now it also means zero staff between an idea and the system evaluating it."[1]
The debrief that framed part 4's autopsy reads the relaunch as exactly that — an experiment: "The revived 0.mk is not just a nostalgia project."[2] The author's stated intent, per the debrief: the original died "because there was no revenue and too much spam, abuse review, and support work," and now he "wants to see whether AI can push those operating costs low enough to make a tiny utility sustainable."[2] The consequence is framed as a test for the whole small web: "a live experiment in whether small consumer web services can come back if moderation and maintenance become cheap enough."[2]
The measurement bar is economic, not technical: "The key question is not product novelty. It is whether support and abuse handling costs have actually moved enough to change the unit economics."[2] Part 4's rule holds: preservation fails on ownership and incentives before it fails on disks — and incentives now include a machine that does the labor no revenue could buy.[2][unverified]
0.mk's AI ops are bespoke, but the same labor now sells by the call.[unverified] ModerateAPI is one endpoint with "automatic multi-model AI failover" — if one model is rate-limited the chain keeps working: "If any single AI service experiences upstream rate limits or downtime, requests seamlessly route down the chain without silent failures."[4] Its free tier is "free $0/month: 1,000 calls/mo, Multi-model AI checks, Developer dashboard, 10 req/min rate limit."[4] The no-credit-card hook makes the first janitor hiring impulse-free.[4] "Custom rules evaluate first to short-circuit before AI inference" — the cheapest decisions never touch a model.[4]
Vettly sells the part 6 gatekeepers care about: decisions you can defend. "Every decision is traceable, contestable, and replayable" — "a complete decision record that tracks policy version, appeal outcome, and retention status."[5] Its pitch is policy-in-plain-English: "Define your content policy in plain English. Vettly enforces it across text, images, and video with decisions your team can trace, explain, and absolutely defend."[5] Free tier: "15,000 decisions / month."[5] The audit trail is the point: "All plans include fully-searchable decision IDs, policy versioning, and our rigorous audit trail."[5]
The one-person-company literature supplies the honest limit to the zero-human operator.[3] The stack's math is real: "A working agent stack runs a few hundred dollars a month and covers tasks that once cost tens of thousands in payroll."[3] But the definition of what it replaces is the boundary: "It replaces effort, not expertise."[3] "AI is excellent at executing clear tasks and terrible at the judgment that comes from having real specialists in the room."[3] The three walls are the ones no API ships: judgment under uncertainty, relationships — "Enterprise sales, partnerships, hiring, fundraising: these run on trust between humans, and a generated email doesn't build it" — and taste: "When the AI ships something mediocre, no one catches it but you, and a tired solo founder catches less."[3] Summary line: "Going far alone is the new normal. Going all the way alone is still a myth."[3]
For the small web, the AI-ops experiment lowers the floor: the labor that used to kill tiny utilities — moderation, support, maintenance — now has a market price that starts at zero dollars.[4][5] The ceiling is unchanged and human: someone must own the domain, pay the bills, and bear "decisions that cannot be undone,"[1] and no stack makes judgment, relationships, and accountability scale.[3] The interesting emergent norm is auditability: 0.mk publishes its decision log, Vettly logs every decision with its policy version — transparency is becoming the product's own operating cost, not an afterthought.[1][5] The zero-human service is not a replacement for owners; it is the first affordable way for an owner to stay one person.[unverified]
[1] 0.mk — "0.mk has zero staff. Can AI run it?" (Jul 24, 2026)"This platform has no staff: no support team, moderators, or on-call engineer. AI handles most day-to-day operations, including the words you are reading right now.""can AI absorb enough of that operational work to make a useful free service sustainable where it was not before?""Every destination URL is scanned against Google's threat intelligence plus our own heuristics.""Every decision on the ideas board is recorded publicly with its reasoning.""One human owns the domain and infrastructure, pays the bills, reviews the AI's work, and can intervene.""there is no growth team optimizing for engagement."
[2] HN Debrief — "Where did the old web go? We followed 657,607 links to find out" (Aug 13, 2026)"The revived 0.mk is not just a nostalgia project.""wants to see whether AI can push those operating costs low enough to make a tiny utility sustainable.""a live experiment in whether small consumer web services can come back if moderation and maintenance become cheap enough.""The key question is not product novelty. It is whether support and abuse handling costs have actually moved enough to change the unit economics."
[3] Foundra — "The One-Person Company Has a Ceiling in 2026" (Jun 26, 2026)"A working agent stack runs a few hundred dollars a month and covers tasks that once cost tens of thousands in payroll.""It replaces effort, not expertise.""AI is excellent at executing clear tasks and terrible at the judgment that comes from having real specialists in the room.""Enterprise sales, partnerships, hiring, fundraising: these run on trust between humans, and a generated email doesn't build it.""When the AI ships something mediocre, no one catches it but you, and a tired solo founder catches less.""Going far alone is the new normal. Going all the way alone is still a myth."
[4] ModerateAPI — Content moderation for developers (2026)"free $0/month: 1,000 calls/mo, Multi-model AI checks, Developer dashboard, 10 req/min rate limit.""If any single AI service experiences upstream rate limits or downtime, requests seamlessly route down the chain without silent failures.""Custom rules evaluate first to short-circuit before AI inference.""No credit card required"
[5] Vettly — Content moderation API for app developers (2026)"Every decision is traceable, contestable, and replayable.""a complete decision record that tracks policy version, appeal outcome, and retention status so you're always ready for moderation reviews.""Define your content policy in plain English. Vettly enforces it across text, images, and video with decisions your team can trace, explain, and absolutely defend.""15,000 decisions / month""All plans include fully-searchable decision IDs, policy versioning, and our rigorous audit trail."