Nine bets on the future of work
Listen to this article 6 min
In September, coding agents wrote 962 of our 973 commits, across 488 releases. Our job now is mostly deciding what to build and checking what comes back. We think most teams will work this way soon. Here are nine bets on what changes.
1. Machines do the execution
Machines already do most of the legwork: drafts, spreadsheets, research, code, follow-ups, reports. And they keep getting better. That frees people for the part that matters most: deciding what's worth doing, who it's for, and whether the result is any good.
We don't know how far this goes. For the next few years it's simple: machines do the work and people make the calls. After that, anyone who claims to know is guessing, us included.
2. Everyone leads
If you run a few machines, you're a manager now, like it or not. Writing a clear brief, giving feedback, knowing when something's done: that used to be the team lead's job. Now it's everyone's. Same with product sense, knowing your customer well enough to say what to build.
ProposalA vague brief used to cost a day. Now it gets you a lot of the wrong thing, fast.
3. Everyone manages themselves
Small teams don't have many managers, so nobody hands you your next task. You spot the problem, fix it and tell the team. A machine will wait for instructions forever. If you wait too, you're doing the machine's job.
What you need instead of a standup is to see what's going on: what was decided, what's in progress and who's on what.
4. Work feels like less, and taste matters most
With machines doing the boring parts, work should feel like less. Less copying numbers between tools, less chasing updates, more time to think. The work gets easier and the decisions get harder.
That's taste, and it comes with a few other things machines don't have: creativity, new ideas, an eye for quality, a feel for what's going on around you, and that X factor some people just have. The best people always had them. Soon they'll be most of the job.
5. Companies get smaller and pay for themselves
Everyone on a team has to stay in touch with everyone else, so the lines between people grow way faster than the team does. Keeping those lines working is a big part of what large companies spend on meetings, managers and status updates.
If a few people and their machines can do what a whole department did, you can stay small. You need less money, raise less or nothing, get profitable sooner, and answer to customers instead of investors. Some companies will still get big. Far fewer will need to.
6. Nobody reads the code
Within a few years, nobody will read code the way developers did in the past. No exceptions. When compilers were new, programmers used to check the machine code the compiler spat out. Today almost nobody does, and nobody thinks that's reckless. Source code is heading the same way.
Picture someone reviewing a 400-line change at 7pm. They skim it, miss the bug on line 312 and approve it. Machines check every line of every change, run every path and try to break it, every single time. As they get better, the human reviewer becomes the riskiest part.
People still decide what the software should do, and whether it does it.
7. Software gets tailored and easy to replace
Software is getting cheap to make, so there'll be a lot more of it, built for very specific needs: one tool for architecture studios, one for clinical trials, one for a five-person agency that works in two languages.
Claude Opus 5.5 · Credits Switching will be easy too: machines will move your data and rebuild your integrations in an afternoon. Lots of teams will just build their own. By the end of September we'd built 29 internal tools just by describing them. With no lock-in, a product has to keep earning its customers every month. Retention is the number that counts.
8. Schools should teach new skills
Most schools still teach the part machines are taking over. They should teach the part above it.
| Schools teach | Work now needs |
|---|---|
| Following instructions | Writing them |
| The expected answer | The right question |
| A grade on the output | Owning the outcome |
| Exams | Critiques |
Design schools have done critiques for decades: you show your work and people tell you what's wrong with it. Every subject could use that. We'll also see new kinds of talent, like a nurse who builds the software their ward needs, or a teenager running a real business. With machines doing the work, one person with good judgment and drive can do what a whole team used to.
9. Companies with the highest AI adoption will dominate
This one's already happening. Companies that don't use AI, or can't get it to stick, are falling behind the ones that do. The gap gets bigger every month.
When factories first got electric motors, many just swapped their steam engine for one big motor and kept everything else. Not much changed. The big wins came later, when they redesigned the whole floor around a small motor at every workstation. Most companies are at step one with AI: a chatbot added to the way they already work.
The ones that rebuild around AI already move faster and do more with fewer people. Every new model makes that lead harder to catch. In most markets, the winner will be whoever uses AI best.
Where this leaves us
These are our bets, and we built WeMachines around them: one place where a small team and its machines work together, and everyone can see what's going on.
The winners will be small teams where everyone leads and machines do the work.
Here's how we work ourselves. Or try it on one project and see what changes in a week.
