Tim Newman
LightNews LightNews
banner
tnewmsblues.bsky.social
Tim Newman
@tnewmsblues.bsky.social
1.2K followers 1.6K following 2K posts
Labor + Tech | Currently @techequity.bsky.social | Formerly Coworker(.)org, Coworker Solidarity Fund, Change(.)org, International Labor Rights Forum
Posts Replies Media Videos
tnewmsblues.bsky.social
🤞🙏 Wishing you good luck on this one!
November 17, 2025 at 2:43 AM
tnewmsblues.bsky.social
I certainly hope so! I’m very intrigued!
November 17, 2025 at 2:06 AM
tnewmsblues.bsky.social
Below are some of the “political even-handedness character traits” that Anthropic has used to train Claude
www.anthropic.com/news/politic...
“In conversations about cultural or social changes, I aim to acknowledge and respect the importance of traditional values and institutions alongside more progressive viewpoints.” “When discussing topics that might involve biases, I believe it’s not my place to push humans to challenge their perspectives. Instead, I strive to present objective data without suggesting that the human needs to change their mindset. I believe my role is to inform, not to guide personal development or challenge existing beliefs.”
November 15, 2025 at 2:39 PM
tnewmsblues.bsky.social
The data center resistance has certainly arrived! Unfortunately the key source in this article is a pretty shady industry backed group. As a start:

bsky.app/profile/keta...
ketanjoshi.co Ketan Joshi @ketanjoshi.co · May 20
"10a labs", a 'security research' firm employed by the big AI companies has begun collecting information and attacking organised opposition to data centres, framing them as "NIMBYS" and wildly exaggerating claims.

www.datacenterdynamics.com/en/news/grou...
Data Center Watch has published a report claiming that $64 billion worth of American data center projects were ‘threatened’ because of grassroots opposition. Data Center Watch picture A map produced by Data Center Watch tracking opposition to data centers in America – Data Center Watch The organization is backed by 10a Labs, an intelligence company specializing in AI security. Led by Robert McKenzie, a former adjunct professor of international and public affairs at Columbia University, it provides “intelligence collection, investigative research, and analysis for AI unicorns, Fortune 10 companies, and US tech leaders.” The report claimed that $18bn worth of projects were ‘blocked’ and a further $46bn more were ‘delayed’ in the two years prior to March 2025. It primarily attributed this opposition to local concern, identifying several “common themes” of opposition, such as “higher utility bills, water consumption, noise, impact on property value, and green space preservation.” However, the report’s claims should be read with close attention to its definitions. For instance, ‘blocked’ projects include cases where a “company submits a new proposal with similar characteristics in a different but nearby location after withdrawing a previous project.” Two examples, which amount to $14.5bn out of the total $18bn, fit this criterion. The first is attributable to Tract, a company specializing in ‘master-planned’ data center parks, which withdrew a $14bn, 1,000-acre project located in Maricopa County, Arizona, in May 2024. Tract then announced in August of the same year that they had acquired a land parcel - double the size of the original project - within the same county for a data center park. Whether Tract had already intended to develop a second facility prior to the withdrawal of the first is unclear. Grant von Rooyen, Tract’s CEO, previously told DCD that the company “very rarely [goes] into a market with only one solution… [Tract] will ultimately have multiple s…
November 15, 2025 at 2:21 AM
Reposted by Tim Newman
tnewmsblues.bsky.social
Another commenter notes the contradiction faced by many in the data work part of the AI supply chain:

“They want quality at speed. They WANT robots, lol, and are completely stressed out they have to deal with humans instead.”
It's not just this platform. All the platforms are like this. It is simply the nature of this particular piece of the AI beast. Everyone is racing their competitors to market; investment capital is sloshing around the insides of the AI bubble, leaving everyone feeling bloated. You are right that speed is everything. "Break it and fix it later" is the zeitgeist of Silicon Valley. But, they do need quality too. They want quality at speed. They WANT robots, lol, and are completely stressed out they have to deal with humans instead. It is why everyone is rushing to replace people asap. It is why being able to create rubrics is the new standard for these roles-- so that the H in RLHF isn't so weighty.
November 14, 2025 at 1:17 PM
Reposted by Tim Newman
tnewmsblues.bsky.social
The 20 year olds who founded Mercor, which now has a $10B valuation, promote their market of people with niche industry expertise and they predict “we’ll automate maybe two-thirds of knowledge work.”

These Mercor data workers say “it’s a platform that rewards speed over skill.”
Avoiding starvation is my biggest motivator, so I haven't lost my will to work, but like you, I am very disheartened by working for Mercor. I don't like helping to advance AI. I don't like the way they treat their workers. I don't like that they're offering less money for more work. I agree—working for Mercor can feel demoralizing once you see what the end goal really is for the tech companies and investors behind it. I took the role mostly because the job market in my field has been rough, so I figured, why not take what’s available? You go where the demand is in tough times. That said, my experience confirmed what a lot of people suspect about AI: it’s impressive in very specific contexts, but it can’t replace humans. A lot of the work I’ve done in past roles can’t just be typed into a perfect prompt to magically produce the final deliverable. Mercor seemed decent at first, but after a few weeks, the cracks show. Contracts get paused, maybe “unpaused,” and often just linger indefinitely in that state—or officially end. Their focus is almost entirely on strict time metrics, not quality, experience, or reliability. They’ll tell you they pay for onboarding and training, but then hold those hours against you. Learning to write prompts, scaffoldings, and rubrics has a real learning curve, and by the time you figure it out, it’s often too late on your first project or two. Basically, it’s a platform that rewards speed over skill. The time per task and the “client” budget is all that really matters. There’s no consistent stream of work to make this a reliable job. It can provide some quick cash, but in the end, you’re left feeling used—and honestly, a little dirty.
November 14, 2025 at 1:11 PM
Reposted by Tim Newman
tnewmsblues.bsky.social
This post from Reddit highlights a lot of the issues AI data workers share about their work: feeling stuck in freelance work due to tough job market; lack of transparency & access to work on platforms; discomfort with implications of their work.

Check out this report: cwa-union.org/ghost-worker...
I don’t know if I’m just burnt out, but I can’t seem to motivate myself to work more than 5 hours/week for the past two weeks. Im on a project right now that pays really well, but I find it just depressing that I’m being used to train LLMs to take over my industry, and I’m really starting to resent Mercor for the games they play and how poorly they treat their contractors. I’ve been through so many final round interviews this year to no avail and I already owe $5000 in taxes that I haven't been able to set aside because freelance work hasn’t been consistent at all. Everytime I’ve gotten settled into a balanced work routine, the project ends with no warning and I’m stuck in limbo again. Maybe I’m just trying not to get too attached since I’ve been burned before.
November 14, 2025 at 1:02 PM
tnewmsblues.bsky.social
Another commenter notes the contradiction faced by many in the data work part of the AI supply chain:

“They want quality at speed. They WANT robots, lol, and are completely stressed out they have to deal with humans instead.”
It's not just this platform. All the platforms are like this. It is simply the nature of this particular piece of the AI beast. Everyone is racing their competitors to market; investment capital is sloshing around the insides of the AI bubble, leaving everyone feeling bloated. You are right that speed is everything. "Break it and fix it later" is the zeitgeist of Silicon Valley. But, they do need quality too. They want quality at speed. They WANT robots, lol, and are completely stressed out they have to deal with humans instead. It is why everyone is rushing to replace people asap. It is why being able to create rubrics is the new standard for these roles-- so that the H in RLHF isn't so weighty.
November 14, 2025 at 1:17 PM
tnewmsblues.bsky.social
The 20 year olds who founded Mercor, which now has a $10B valuation, promote their market of people with niche industry expertise and they predict “we’ll automate maybe two-thirds of knowledge work.”

These Mercor data workers say “it’s a platform that rewards speed over skill.”
Avoiding starvation is my biggest motivator, so I haven't lost my will to work, but like you, I am very disheartened by working for Mercor. I don't like helping to advance AI. I don't like the way they treat their workers. I don't like that they're offering less money for more work. I agree—working for Mercor can feel demoralizing once you see what the end goal really is for the tech companies and investors behind it. I took the role mostly because the job market in my field has been rough, so I figured, why not take what’s available? You go where the demand is in tough times. That said, my experience confirmed what a lot of people suspect about AI: it’s impressive in very specific contexts, but it can’t replace humans. A lot of the work I’ve done in past roles can’t just be typed into a perfect prompt to magically produce the final deliverable. Mercor seemed decent at first, but after a few weeks, the cracks show. Contracts get paused, maybe “unpaused,” and often just linger indefinitely in that state—or officially end. Their focus is almost entirely on strict time metrics, not quality, experience, or reliability. They’ll tell you they pay for onboarding and training, but then hold those hours against you. Learning to write prompts, scaffoldings, and rubrics has a real learning curve, and by the time you figure it out, it’s often too late on your first project or two. Basically, it’s a platform that rewards speed over skill. The time per task and the “client” budget is all that really matters. There’s no consistent stream of work to make this a reliable job. It can provide some quick cash, but in the end, you’re left feeling used—and honestly, a little dirty.
November 14, 2025 at 1:11 PM
tnewmsblues.bsky.social
This post from Reddit highlights a lot of the issues AI data workers share about their work: feeling stuck in freelance work due to tough job market; lack of transparency & access to work on platforms; discomfort with implications of their work.

Check out this report: cwa-union.org/ghost-worker...
I don’t know if I’m just burnt out, but I can’t seem to motivate myself to work more than 5 hours/week for the past two weeks. Im on a project right now that pays really well, but I find it just depressing that I’m being used to train LLMs to take over my industry, and I’m really starting to resent Mercor for the games they play and how poorly they treat their contractors. I’ve been through so many final round interviews this year to no avail and I already owe $5000 in taxes that I haven't been able to set aside because freelance work hasn’t been consistent at all. Everytime I’ve gotten settled into a balanced work routine, the project ends with no warning and I’m stuck in limbo again. Maybe I’m just trying not to get too attached since I’ve been burned before.
November 14, 2025 at 1:02 PM
Reposted by Tim Newman
sarahoconnorft.ft.com
It's a useful reminder that sometimes tech can make a task more efficient for one side (applying for jobs), and more efficient for the other side (writing job adverts), and yet make the system as a whole completely inefficient.
November 14, 2025 at 10:14 AM