Breaking Even — April 16, 2026
Handshake AI's projects are dying in real time, Babel Audio is the sleeper nobody expected, and the Mercor silence is deafening.
Week of April 9 – April 16, 2026
I tracked 866 posts across nine platforms this week. Market health score: Weak. Same word as last week. Same feeling, too — except this time I watched one of the highest-paying platforms in the space start losing projects in real time while a voice-chat app nobody was talking about three months ago quietly became the only platform generating genuinely happy posts. This market doesn't do subtlety anymore.
Handshake AI — The Pay Cliff Arrives
I've been saying for weeks that Handshake AI is both the highest-paying and most precarious platform in the space. This week proved it.
Project O ended on April 14. No warning. No transition plan. Workers who were earning $80/hr woke up to empty dashboards. Project H — the other major earner — is down to roughly three available tasks at any given time. Workers are posting about being stuck in "quality tiering" pauses for over a month with no recourse and no work to submit to prove they've improved. One person described being paused on March 10th, following up on the 19th, getting a template response, then hearing absolutely nothing through April 15th. Over a month of silence while being told to "consistently submit high-quality work" — with no work to submit.
The sentiment score sits at 45/100, flat. But flat doesn't capture what's happening underneath. Payment processing failures showed up this week — workers posting about failed payment notifications. The weekly earnings cap is being applied to projects that were previously unlimited. And the offboarding vocabulary keeps growing: "I got hit with a weekly limit," "discouraged," "it's a sad day."
Here's what I keep coming back to: Handshake told me, specifically, that they would provide feedback when workers underperformed. That we wouldn't get silently removed like what happens on Outlier. That was the pitch. That was the differentiator. And right now, a buddy of mine is on his third follow-up asking for the feedback he was promised, getting nothing but generic responses that don't address what he actually asked.
I wrote a full Handshake AI review this week because the platform deserves more than a newsletter section. The short version: the $80/hr is real, the instability is real, and the fall from $80/hr to $0/hr breaks bones if you forgot the rules.
Status: Critical | Sentiment: 45 | Rates: $60–$100/hr | Hiring: Active but narrowing
Outlier AI — Whitebeard Hits the Wall
Last week it was Aether bleeding out. This week, the replacement is already showing cracks.
Whitebeard — the project Outlier funneled Aether refugees into — hit daily task limits almost immediately. Workers are posting about reaching their cap and sitting idle. The "Unified Skill Screening" emails went out to a batch of workers this week, but here's the kicker: the screening doesn't appear on their dashboards yet. You get an email saying you've been selected for something, log in, and see nothing. Classic Outlier.
The Aether 500 errors continue. Workers are still getting server errors when trying to load tasks. The AirTM situation from last week — someone with $1,023 trapped in a banned account — remains unresolved. And the account ban rate hit 12.3% of all posts. Let me say that again: more than one in eight Outlier posts this week is someone getting banned.
The 293 posts we analyzed tell the same story they've been telling for months, just louder. Hubstaff disputes are still the second-biggest complaint — 30 posts about screenshot tracking disagreements and the question nobody at Outlier will officially answer: do you start the paid timer during mandatory tutorials?
If you're still on Outlier, the move is Whitebeard if you can get past the task limits, or start diversifying yesterday. The generalist era at Scale AI's crown jewel is closing, and the door is not reopening.
Status: Warning | Sentiment: 38 (flat) | Rates: $15–$50/hr | Ban rate: 12.3%
Babel Audio — The Sleeper Hit of 2026
This is the story I didn't expect to be writing.
Babel Audio hit a 3x volume spike this week — 103 posts versus a 34.9 average. And here's what makes it remarkable: almost all of it was positive. No payment failures. No ban reports. No offboarding horror stories. Just people saying they like the work and they're getting paid.
A worker who started on March 12th posted a glowing one-month review. Two comprehensive guides showed up on beermoney forums calling it one of the best AI gig options available. A DataAnnotation worker specifically migrated to Babel Audio and posted about the transition. English-Video2 and Video2-Solo channels are confirmed active. English-Generalist is hitting a 403 capacity cap — meaning they have too many people wanting to work, not too few tasks.
The pay isn't Handshake money — we're looking at roughly $17/hr for the conversational work ($4.38 per 15-minute session). But the consistency is there. The work is steady. The vibes are, against all odds in this industry, good.
The catch: your workspace needs to be basically soundproof. No keyboard clicks, no heavy breathing, no house noise. If you have kids, dogs, or a neighbor who mows at 7am, you'll need to sort that out.
This is rare positive buzz in a space dominated by complaints. I don't know how long it lasts. Nothing lasts in this industry. But right now, Babel Audio is the platform people are talking about without anger in their voice, and that counts for something.
Status: Operational | Sentiment: 59 | Rates: ~$17/hr | Hiring: Active (capacity limited)
Alignerr — Still Falling
Alignerr's sentiment dropped to 27/100 this week — the lowest of any platform we track, and still trending down. That's not a floor. That's a platform in free fall.
The CHP CLAUDE / Transcript / Coding project is the epicenter. Reviewers are retroactively failing previously-approved tasks. Not reviewing new work — going back and flipping passed tasks to fails. Roughly 500 workers are on this project. The community response has been unified frustration. Empty queues are showing up in 15.7% of posts. Payment issues in 13.7%.
The LMArena V2 Discord server vanished this week — taking work history with it. If you were holding off on submitting more prompts until reviews came back, your account is now staring at empty space where your work used to be.
Hiring status has been downgraded to Caution. When the most-discussed topic on a platform's subreddit is the name of the parent company (Labelbox), and not the actual work, something structural is breaking.
Status: Volatile | Sentiment: 27 (↓) | Rates: $15–$60/hr | Hiring: Caution
DataAnnotation — The Calm in the Storm
DataAnnotation continues to be the most stable platform in the space. Sentiment: 57, trending up. Status: Operational. The subreddit reads like a different industry entirely.
People are discovering the save progress button. Workers are comparing notes on adversarial tasks — one person wrote "Yes please pay me to say bad stuff to the robot!" Bilingual workers are comparing dashboards and, unsurprisingly, finding theirs fuller than everyone else's. Language specialist work appeared in 11% of all posts this week.
The four-month waitlist story keeps repeating: someone who applied as a generalist in December 2025 finally got their acceptance this week. If you applied and haven't heard back, don't assume you failed. The pipeline is just that slow.
Rates are holding at $28–$65/hr with a $47 average — solid middle of the market. No mass offboardings. No retroactive task failures. No disappearing Discord servers. In this market, boring is a feature.
Status: Operational | Sentiment: 57 (↑) | Rates: $28–$65/hr | Hiring: Active (long waitlist)
Mercor — The Silence After the Fire
Zero new Reddit posts. The community has gone completely silent, which — if you've been following this — is somehow more unsettling than the five lawsuits filed in a single week.
TechCrunch ran a piece titled "After data breach, $10B-valued startup Mercor is having a month." Business Insider detailed how contractor NaTivia Esson's lawsuit alleges she worked for Mercor from March 2025 to March 2026 before the breach exposed her data. HR Dive reported that plaintiffs alleged Mercor failed to train staff on basic cybersecurity practices. Staffing Industry Analysts covered the class action filings.
Meta's indefinite pause on all Mercor contracts remains in effect. The lead lawsuit — Gill v. Mercor.io Corporation — is proceeding as a proposed nationwide class action.
If you're a Mercor contractor: assume your data was exposed. Freeze your credit. Change your passwords. We're maintaining the Mercor platform page as this develops.
Status: Operational (legally compromised) | Sentiment: 72 (pre-breach baseline) | Hiring: Paused by major clients
The Rest of the Field
Telus AI — Sentiment: 35, flat. Empty queues remain the #1 complaint at 17.3% of posts. Named a Leader in Everest Group's PEAK Matrix for data labeling, and their AI Community now exceeds 1 million annotators. The corporate side is thriving. The individual worker side is dry. Rates: $10–$30/hr.
Mindrift — Recovered to Operational. Sentiment: 55, trending up. Task quality scoring is the main conversation (21% of posts). Quietly stable. Rates: $20–$45/hr.
Stellar AI — Sentiment: 48, trending down. Technical issues are the theme — loading errors, site outages, broken project pages. One project (EAIAIWT) has workers reporting 6+ hour tasks going unpaid. Operational but wobbly. Rates: $18–$25/hr.
The Bigger Picture
The market composite sits at 48/100 this week. Up slightly from 45 last week. Still weak. Still reorganizing.
Here's what I see when I look at these numbers together: the industry is sorting itself into lanes, and it's doing it fast. Handshake AI pays $80/hr and just lost two projects. Babel Audio pays $17/hr and has a capacity cap from too many people wanting in. DataAnnotation is stable and boring at $47/hr average. Alignerr is in free fall. Outlier is capping the replacement for the project they just killed.
The platforms that are surviving are the ones where the work is specific enough that they can't just swap you out overnight. Voice data. Bilingual expertise. Domain credentials. The generalist "rate this AI response 1-5" work is either gone or paying so little it's barely worth tracking.
If you're still a generalist in this space, this is the week to look hard at what skills you actually have that a platform can't easily replace. Because the middle is getting squeezed, and the squeeze isn't stopping.
Track all platform sentiment on the Dashboard. Compare platforms side-by-side on our Platforms page. New this week: Handshake AI — The $80/hr Platform That Keeps Offboarding People.
Get this in your inbox on the 7th of every month — free.
Subscribe to In Hindsight →