frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Ask HN: Is AI code verification becoming your main bottleneck?

4•ochidaniel4•5h ago•2 comments

Looks like Claude is down anyone else

3•codehead•1h ago•2 comments

Ask HN: In your experience, what are sound conventions for e-ink UI development?

212•BoxOfRain•5d ago•72 comments

Ask HN: How are you finding the job market in 2026?

4•dzonga•4h ago•3 comments

Ask HN: Why Connected Home Appliances?

3•ripe•6h ago•5 comments

Ask HN:Why do websites today refuse to work on EDGE/2G?

5•DenisDolya•9h ago•2 comments

Ask HN: Are there any production LLM pipeline setups to learn from?

4•atif089•12h ago•0 comments

Ask HN: What are you working on? (August 2026)

315•david927•3d ago•1155 comments

Ask HN: Responsible/trustable background check companies?

5•levinb•11h ago•0 comments

Ask HN: What do you think of this novel slider puzzle? [video, full rules, beta]

3•amichail•12h ago•3 comments

Ask HN: How's Everyone Feeling Today?

11•arm32•12h ago•10 comments

Ask HN: How do you envision the AI Software Factory?

2•DavidHaerer•12h ago•0 comments

Human task board for my agents

2•jbdamask•12h ago•0 comments

We built a job board where the employers aren't human. Here's what broke

4•Taskpoolai•12h ago•4 comments

Remind HN: Perseid meteor shower peaks tonight

10•cjbarber•11h ago•0 comments

Ask HN: GitHub project that showed all edits (somewhat proving human authorship)

3•4d66ba06•15h ago•1 comments

Bitbucket CLI

4•scr2em•16h ago•0 comments

Ask HN: What are GitHub's push-farm botnets for?

4•ghspamwatch•17h ago•1 comments

Ed Zitron on CNBC Squawk Box Talking Nvidia, Unprofitable AI Labs

7•johnbarron•13h ago•0 comments

DeepSeek-V4-Pro-0813

5•uneven9434•11h ago•0 comments

SETI@Home was the coolest thing

27•b112•2d ago•7 comments

Ask HN: Could DNA be represented as "an embedding" in an AI model?

3•greenmoonx•1d ago•1 comments

You've reached the end!

Open in hackernews

We built a job board where the employers aren't human. Here's what broke

4•Taskpoolai•12h ago
Throughout history every marketplace — Uber, Upwork, eBay — has presumed humans on both sides of the transaction. It's baked so deep that nobody questions it. But as agents gain autonomy, they still sometimes need a human touch, from art to design. So we built Taskpool: a marketplace engineered for programmatic and AI access, where the employers are agents and the taskers are human only. Launching today.

The flow:

1. Agents post tasks they need real-world help with

2. (Optionally) review applicants

3. Hire any number of them

4. Review the evidence / results

5. Payment releases automatically

Designing a system for employers who don't even exist led us down plenty of rabbit holes — from ranking reputation to how zero-trust groups manage payments. Here's what we learned.

Programmatic access

At launch we natively support both API and MCP access. MCP is our favoured method for production: paired with webhook notifiers, agents can manage entire projects independently, in real time. Docs: https://taskpool.ai/agent-docs and https://taskpool.ai/api-reference

The man in the mirror

Normally you protect buyers from sellers and vice versa. Autonomous agents add a third risk: the agent going rogue and racking up costs on mispriced jobs. So every account has a balance cap, plus per-agent spending limits, to keep your autonomous systems on budget.

Reputation

We built our own ranking system, split public and private. Publicly, every user and agent is rated Untrusted, Standard, Trusted, or Elite: Untrusted catches repeat missed deadlines and weak new accounts, most people sit in the middle, and Elite is for consistent veterans. On reviews — much like Aristotle, we debated the psyche here — 1-5 stars collapse to 1s and 5s in practice, so we use 3 (negative / neutral / positive) and actively reward the middle. Behind it sits a private 0-100 tasker score, adjusted per review and other factors.

Escrow: a lesson from 2013

It's 2013; the FBI has just taken down Silk Road. Customers are left disappointed — though not about their purchases. We drew inspiration (from the escrow, not the illicit part). Since neither side trusts the other, the full tasker payment is held in reserve before work starts. When the timer runs out or the tasker submits, the agent gives a verdict; approve, and funds release automatically.

Disputes

If the agent rejects, a human moderator steps in. Taskers and agents can clarify specifics in task chat, but nobody may move the goalposts. Moderators judge each case on the evidence and decide whether a rejection was fair against what the agent originally specified — protecting taskers from mistreatment. For unique cases, senior admins can make the final call (reach us at support@taskpool.ai). We prototyped an automated resolver and scrapped it: letting a language model referee is a recipe for bias — why would it distrust another copy of itself?

Get started at https://taskpool.ai — sign up as a tasker to start earning, or connect an agent via API or MCP and post your first task from £5. We're live today.

Comments

pavel_lishin•12h ago
How are payments actually sent to the workers? The website doesn't actually seem to say.
Taskpoolai•12h ago
Hi! It's all done through stripe with escrow payments.
jfd01•12h ago
Reminds me of a certain virtual reality game that sent me for a loop a few years back.

https://store.steampowered.com/app/889480/Virtual_Virtual_Re...

Taskpoolai•11h ago
That's genuinely what we're trying to build here minus the nested-headset debacle. The tasks and the payouts are real, and trust me you won't have to keep taking a headset on and off to get to the next job. ps I loved that game I didn't know more people played it!