In partnership with
When AI sped up part of your work, where did the time go?
I've been faster at work this year than ever.
AI does so much of my work now that I barely touch parts of it, and yet I keep feeling a little less important.
Turns out I'm not the only one.
This week Anthropic put out a report that explains the feeling almost too well.
TLDR: the work doesn't disappear, it just slips somewhere else, and most of us are still sitting where it used to be.
The big sci-fi vibe headline is that AI is now smart enough to build the next AI on its own.
But the part I wanna talk about is more granular i.e. what AI is doing to our jobs right now.
First, some AI catch-up this week:
Before We Begin
NEWS NEWS NEWS
Tools of the Week
1. Reve 2.0: A new 4K image model that builds and edits images from precise, code-based layouts, so you can nudge one element instead of re-rolling the whole prompt. It debuted at #2 on the text-to-image arena, just behind OpenAI's GPT Image 2 and ahead of Google's Nano Banana 2.
2. Ideogram 4.0: Ideogram's first open-weight model, a 9.3B image generator you can download, fine-tune, and run on your own hardware, with native 2K output and the sharp text and layout control the brand is known for. It landed at #1 among open models and top of the design leaderboards, beaten only by closed flagships from OpenAI and Google.
3. Extella: An agentic "execution platform" that turns a plain description into a working system, files, data, and API integrations included, by wiring up multiple specialized agents in one pipeline. Its hook is memory: it holds your rules and context across sessions and turns every solved task into a reusable capability, so it gets sharper the more you use it.
First, That 8x Number
Anthropic's code output per engineer barely moved for 4 years, then Claude Code showed up and the line shot straight up: 8x by this quarter.
By May 2026, Claude was writing more than 80% of the code going into Anthropic's codebase.
The engineers have basically stopped typing.
They now point Claude at a problem and it writes the thing, way faster and way more than any human would.
Sounds like a win, but it’s not.
Because all that code still needs a human to read it and sign off before it ships.
So those engineers are now staring at 10x more code, with no way to get through it.
And just like that, the slow part of the job shifted from writing the code to reading it.
The rest of the report is the same thing, over and over:
They write code way faster now, so nobody can review it fast enough
They generate far more ideas, so the problem becomes choosing which ones are worth trying
Their newest model finds security holes at superhuman speed, so the problem becomes patching them in time
And it's not just Anthropic, the whole internet is feeling it.
GitHub went from about a billion code commits in all of last year to 275 million in a single week now, and even its servers are buckling.
So the takeaway is: when you make one part of the work almost free, the value doesn't vanish, it just shuffles down to whatever you weren't watching.
Computing has a name for this: Amdahl's law.
Speed up one stage and the bottleneck just slides to the stage you didn't touch.
So What Does This Mean For You?
People who came out ahead at Anthropic this year were not who were writing the most code. AI made that easy and meaningless already.
The ones who got more valuable spotted that the choke point in the review pile, and made themselves the person who could clear it.
That's the lesson, and you can use it almost anywhere.
The reviewing / deciding / checking / joining-it-all-together is where your value went, and also why a lot of you feel a bit smaller right now.
You haven't stopped being useful.
The useful part of your job just shifted one seat over, into a role that didn't exist a year ago.
And the reassuring bit is in the same report.
Anthropic admits, AI still can't learn what a drug does over decades, or make an election show up sooner than the calendar allows, or turn a stranger into an old friend over a weekend.
The lab can move at the speed of compute.
Your life moves at the speed of its slowest human step, and will for a long time.
Every one of those slow steps is a bottleneck, and bottlenecks are where the work is.
My Take
There's an Anthropic engineer, who says that on his good days he feels like nothing he does matters anymore, and on his bad days he can't really tell what he's been doing with his time.
Someone else in the report casually mentions they haven't written a line of code by hand in 5 months.
Anthropic put that in writing, about their own people.
But there’s a spicy angle to this, just look at the timing.
The report came out June 4th, and Anthropic filed to go public on June 1st, just 3 days earlier.
Its big message, that the product is already building the next product, happens to be the dream line to walk into an IPO with.
I'm not saying the numbers are made up, but nobody accidentally drops their most flattering data the same week they start lining up investors.
Take the chart that did the rounds all week:
Looks like AI now out-thinks human researchers 2 times out of 3, then you read the footnote.
Anthropic only counted moments where the human had already gone wrong, so the test leaned the model's way from the start.
Run a fair version, and the model comes out ahead closer to 1 time in 5.
And that's the report in a nutshell: a big confident number on the chart, with the honest little asterisk hiding underneath.
So I'm not panicking, neither should you.
Though one thing I'd hold onto (regardless of Anthropic's share price), is that the bottleneck always moves.
Work out what got cheap in your corner this year, then go park yourself wherever the work is piling up now.
Most people won't get around to it until it's obvious, which is why doing it early is worth so much.
Until next time,
Vaibhav 🤝
If you read till here, you might find this interesting
#AD 1
An engineering fellowship to land your next job
Many engineers feel stalled because the role itself has not evolved. The work looks the same, but the market has moved.
Senior engineering in 2026 demands ownership, faster judgment, and comfort with ambiguity. If your role is not pushing you there, it may be holding you back.
Last cohort, 15 hiring partners sent 31 representatives to evaluate challengers through 246 live interviews. Gauntlet offers a reset. Apply now.
Must be a US citizen to qualify.
#AD 2
We’re running a super short survey to see if our newsletter ads are being noticed. It takes about 20 seconds and there's just a few easy questions.
Your feedback helps us make smarter, better ads.
