In partnership with
Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
Hi {{first_name|everyone}},
Anthropic just looked at its own workforce, and one of the odd findings is that a fifth of employees say their jobs have shifted from doing the work to managing Claudes, with one engineer calling it babysitting.
Next, we’ve got a few new tools, a couple of quick updates on OpenAI and DeepSeek, and a look at AI-assisted oil paintings now hanging in a New York gallery.
Let’s see what’s inside!
🤿 DEEP DIVE
Anthropic's AI Productivity Paradox
More output, fewer junior hires, and a whole lot of babysitting.
Apparently, one of the problems with having AI do more of your work is that you end up spending a lot of time managing the AI.
At least that's what seems to be happening inside Anthropic.
The company just published a report looking at how its own workforce has changed over the past year. And since Anthropic has built so much of its daily work around Claude, it's a pretty interesting look at what happens when AI becomes part of almost everything a company does.
And some of the numbers are quite telling.
Fewer juniors, more AI
Since early 2025, the proportion of junior-level employees in Anthropic's technical departments has dropped from 21% to 14%.
But the hiring numbers are even more striking. Junior employees went from making up 27% of technical hires to just 9%.
Meanwhile, technical teams themselves have shrunk as a share of the overall workforce, from 61% to 56%.
Anthropic says part of that is simply the business growing and needing more people in other departments. But there's another explanation too. Its coding tools are increasingly allowing non-technical employees to build software themselves.
So it's not just that engineers are using AI to write more code. People who aren't engineers are apparently writing software too.
The babysitting part
This is where things get a little funny.
In an April 2026 employee survey, 27% said Claude had made them more productive. But 20% said their jobs had shifted from actually doing work to managing Claudes.
One engineer even compared the experience to babysitting.
Basically, every Claude you're managing has something that needs your attention, and they all seem to think their concerns are equally important.
Which gets at something Anthropic acknowledges in the report.
Claude can produce so much work that human attention and judgment are becoming the bottleneck. Someone still has to read everything, organize it, make decisions, and figure out what's actually useful.
And apparently there's an "eternal pressure" inside the company to move as fast as possible and use Claude for everything.
More work done, but more work to check
To be fair, Anthropic says the productivity gains are showing up in actual business activity.
Employees are producing more code per person, sales reps are closing more sales, and the company is processing more contracts.
It also hasn't found evidence that code quality is declining, although it acknowledges that producing more code doesn't automatically mean producing better code or better business results.
I think that's the part worth remembering.
Anthropic is getting more output from its workforce while hiring proportionally fewer junior technical employees. But at the same time, some employees are spending less time doing the work themselves and more time reviewing what AI produces.
So the work isn't necessarily disappearing. Some of it is changing from doing things to managing things that do them.
And at Anthropic, that apparently includes a surprising amount of babysitting.
The quick take
Junior hiring is falling sharply. Junior employees now account for just 9% of technical hires, down from 27%.
Technical teams make up less of the company. Their workforce share has fallen from 61% to 56%, partly as non-technical employees use AI coding tools.
AI management is becoming part of the job. One in five surveyed employees said their work had shifted toward managing Claudes.
Human attention is the bottleneck. Anthropic reports higher productivity, but reviewing and organizing AI-generated work still takes time.
🤝 FROM OUR FRIENDS
A quick share we thought you might find useful
Nobody's coming to approve your idea
People assume that real businesses get built by other people with more resources. It's how great ideas get stuck in notes apps and conversations.
But you don't need six months and a developer's salary to validate your concept. You just need a weekend. Describe what you want to exist and Lovable builds it: a real site, a real product, ready to accept real payments from real customers. Not a mockup to show people. An actual thing they can use.
You don't need anyone's sign-off to get started. Take $5 off your first month.
Discount of $5 applies to first month of any paid plan. You need to sign up via this webpage in order to claim the discount. The discount applies to new and returning users without an active paid plan. Terms apply.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
Get a personal assistant that handles tasks right from WhatsApp.
Why it’s useful: Handy when you've got too much on your plate and want someone to keep track of projects and get things done without juggling more apps.
🧠 Beam
An AI model built to tackle complex coding and reasoning tasks.
Why it’s useful: Worth a look for developers building AI agents or working on problems that need more than simple text generation.
💻 Beam.win
Run your AI agents on one Mac, control them from another.
Why it’s useful: Useful for developers who want to keep agents and builds running on a dedicated Mac while accessing their windows from their everyday machine.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
AI-Generated Industrial Worlds Become Real Oil Paintings
Artist Ai Makita uses Midjourney and other digital techniques to imagine intricate networks of pipes and machinery, then recreates the designs as physical oil paintings. The results look like impossible industrial worlds, and several are currently on display at Fou Gallery in New York.
AI used:
Midjourney. Photoshop is also confirmed in the workflow. The gallery specifies that only some paintings were developed with Midjourney, without identifying every individual AI-assisted piece.
👀 ON OUR RADAR
One more thing we thought you'd like
What’s still sitting on your to-do list?
You don’t need an AI team to build an AI agent. With Skydive, the person who knows the work can create the agent to do it.
Start an agent off with a role, teach them how your team operates, connect your tools, and put them to work.
⚡ QUICK HITS
The latest in AI, tech, and productivity worth knowing
OpenAI has expanded its Chrome extension so ChatGPT can read your open tabs, compare information across websites, and even fill out forms. In ChatGPT Work, it goes further, handling tasks like sending messages, downloading files, and organizing tabs. The extension works across Chromium-based browsers and is available on all ChatGPT plans.
The Department of Homeland Security has banned DeepSeek and technology built using its AI models from most civilian federal agencies. The order comes amid growing concerns over Chinese AI, although officials haven't detailed the specific security risks. What's interesting is that DeepSeek's cheaper, open-source models are already used by American tech companies, including Snowflake and Vercel.
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
