'I'm Doing 10x the Work for €40 a Month.' 'Are You Making 10x More Income?' 'Not Yet.'

In a thread about companies pulling back on AI spending, spiderfarmer described their setup: "Never spent more than 40 euros per month on the base plans for Claude and OpenAI. And I'm doing 10x the amount of work I did before."

6stringmerc asked the only question that matters: "Are you making 10x more income for yourself?"

The answer: "Not yet. But I finished all of my sideprojects in addition to my regular work and can now focus on doing more consultancy jobs again."

That exchange is three comments long and it contains the whole story that two Hacker News threads spent 379 comments circling. Output is up, sometimes enormously. Income moved separately, or not at all, and the reason has nothing to do with how good the models got.

The numbers people actually cited

The bigger thread was about a Stanford policy brief on what is happening to jobs. I could not open the brief itself, so what follows is what commenters brought to the discussion, and I am flagging that rather than pretending to have read the source.

keeda contributed the adoption figures, with the caveat attached: "even though 50%+ of Americans currently self-report (major caveat) using AI at least weekly, they use it for only 6% of work hours." A Google study they cited calls this "broad but shallow use." The same surveys find time savings of 2% of working hours, which keeda works out to "a whopping 30%+ productivity boost per hour of AI used."

Both halves matter. Per hour of use, the gains look real. Across the working week, the total is 2%, because almost nobody is using it for most of what they do. keeda's own worry is that this is the early part of a curve: "I fear the impact will happen gradually, as adoption inches up... and then suddenly."

jdlshore brought the study that gets quoted in every one of these arguments: "That early 2025 METR study was particularly interesting because participants self-evaluated themselves as 20% faster, but the measurements showed they were actually 19% slower." Then the detail that usually gets dropped: "Unfortunately, METR hasn't been able to replicate the study because they couldn't find enough willing participants."

A study nobody could repeat is not a settled finding. It is also the only controlled measurement of that gap anyone in the thread could point to, and the gap it found runs in the direction people find least flattering.

The argument about timing, which is genuinely unresolved

simonw made the standard objection to any study of this: "coding agents (Claude Code, OpenAI Codex) only started working really well in late November, which for most people meant early January due to the December break. Studies that mainly focus on 2022 to end of 2025 might be missing out on a material uptick in capabilities."

Several people independently reported the same timeline. qarl2: "For the first time last December I noticed the harnesses performing like actual workers. Everything impressive has happened in the last six months."

jcranmer named the problem with using that as a defence, and did it well enough that I want to quote the whole thing: "every study that comes out saying 'no, that wasn't,' the response to that is, 'no, there's a new change that is totally the key inflection point!' Or, put differently: if a study comes out next year saying they don't see major impacts from AI in 2026, will you admit your viewpoint as being wrong, or is your response going to be 'no, there was a massive inflection point in September 2026 that completely invalidates the paper'?"

simonw's reply: "Things are allowed to get better more than once!"

Both are right, which is why the thread never lands. Research on a fast-moving field is always measuring the past. A hypothesis that can absorb every negative result is also not doing much work. If you are trying to decide what to do with your own week, neither position gives you a number.

Layoffs, and what they were actually for

AbsurdCensor offered the deflationary read: "programmers and IT were severely overhired during the pandemic, there are massive job losses now, and it's easy to blame AI when in reality there are a lot of economic factors."

samstokes pointed at layoffs.fyi and noted the ambiguity that comes with it: an uptick starting in 2026 "could be explained either by AI actually causing more layoffs, or by AI becoming an even better excuse for layoffs."

overgard supplied a single case with the timeline attached: laid off from a big tech company at the end of January with "AI" as the stated reason, new job by April without much effort, and the hiring company reporting that the search had taken a long time. Their conclusion: "I think narrative of job loss is way overblown."

One person's four months is not a labour market. It is still more evidence than most of the predictions in the thread carry.

Will AI take my job, and why that is the wrong question for most people here

If you sell your hours to an employer, the question makes sense and the answer depends on your tasks, not your title. chewbacha described where the gains land: "it's concentrated in lesser experienced engineers and vanishes or goes negative with highly experienced engineers. LLMs move your baseline towards the mean. If you are below average it improves and if you are above it hinders."

The hiring side of that showed up in the same thread from two directions. ralusek: "Now, I don't want any additional engineers... the prospect of having junior engineers using AI is absolutely terrifying to me. I can't eyeball their code to get a sense of how good of an engineer they are anymore." VladVladikoff, running a small company, reached the same place: "I am no longer interested in hiring any juniors, as all they do is copy and paste LLM output without thinking, which anyone could do."

That is two managers describing the same decision from opposite premises about what AI does to quality. The junior role gets squeezed either way.

But if you are trying to earn on your own account (freelancing, selling a product, building an audience), the exposure question is secondary. Your problem is the one in spiderfarmer's answer. You can now produce far more, and nothing about producing more creates a buyer.

pjmlp said it in one line, about agency work: "AI empowers to do even more with even less people, however there isn't enough project demand to keep everyone busy."

Demand is the constraint. It was the constraint before the models got good, and the models do not touch it. We went through the same arithmetic in revenue versus profit and again in realistic passive income: the number that moves your income is almost never the one that measures your output.

Where the money in these threads actually was

consumer451 drew the line that is worth writing down: "If you sell 'AI' maybe, if you sell products that happen to use LLMs to provide services previously not possible, the money still exists in my experience."

Then, on watching competitors deploy badly: "What I keep saying in internal meetings is: 'I am so glad these people are this bad at deploying these tools.' It really leaves the door open for folks like us."

That is the same position we keep arriving at from different threads. The people making money are not selling the capability. They are selling something a customer already wanted, produced more cheaply than before. thraway3837, in a different thread we covered, replaced a $120,000-a-year platform with under $1,000 a month of subscriptions. The saving is the product. Nobody paid them for the AI.

And spiderfarmer's answer, read again, is not a failure. The hours went into finished side projects and freed capacity for consultancy work, which is a plan for converting output into income. It just runs through clients, not through production. Income was flat at the moment they were asked because the conversion step had not happened yet.

What the threads do not establish

No income figures. Across 379 comments about whether AI is changing work and money, almost nobody posted a before and after on their own earnings. The one person who was asked directly said "not yet."

No resolution on productivity. The controlled study says slower, the self-reports say faster, the study could not be replicated, and the timing objection to it is legitimate.

No agreement on the layoffs. Overhiring and AI both explain the same curve, and the people invoking each are reading the same charts.

What survives all of it: adoption is shallow (6% of work hours), the per-hour gains where people do use it look real, demand did not increase, and the junior rung of the ladder is getting pulled up by managers who disagree about why.

The part you can act on this week

If your income has not moved while your output has, the broken link is downstream of you, and finding it takes an hour rather than a strategy.

The output-income-gap prompt in the panel writes out every step between you making something and money arriving, then names the one step that does not scale with production. Usually it is demand, sometimes it is your pay structure, occasionally it is pricing you have not touched in two years. demand-before-supply tests whether anyone wants the extra thing you can now make, with a kill criterion you define before you start so you cannot argue yourself out of it later. freed-hours-allocation deals with the hours themselves, which otherwise get quietly absorbed into more unpaid output. And role-exposure-check is the "will AI take my job" question asked task by task, with the honest answer about how much of your week sits in the automatable bucket.

The threads spent 379 comments on whether the machines are coming for the work. The more useful question, for anyone selling their own output, is the one 6stringmerc asked in eight words, and it is worth asking yourself every quarter until the answer changes.