AI revenue up 700% year over year, but grunt-work tasks still 6x behind humans
The 2026 evidence on AI progress: revenue exploding, coding agents maturing, but grunt-work tasks still lag; timelines shortened by about a year, though the long run remains uncertain.
The video won't play here. Listen to the audio instead:
The argument · timestamps estimated from transcript position
The reversal isn't a new breakthrough — scaling still works
In October 2025 AI sentiment was pessimistic: GPT-5 was seen as a failure and agents as unreliable. Two months later Claude Opus 4.5 changed everything, programmers found agents genuinely usable, and figures like Karpathy shifted their stance. The key isn't breakthrough technology but that scaling and RL keep working: as models get smarter, error rates on long-chain tasks fall. The Epoch Capabilities Index shows that after reasoning models appeared in September 2024, the pace of progress accelerated 2-3x and has held there since.
— Rob WiblinGross margins rose to 70% — this isn't buying revenue at a loss
OpenAI and Anthropic grew annualised revenue 700% over the past six months, and 1600% over the past three. Anthropic grew annualised revenue 8400% over the past 5.5 months, reaching a $47 billion annualised run rate in May. Gross margins rose from 38% to above 70%, showing this isn't loss-making sales. Revenue growth proves real value, investors keep putting money in, and the industry is in a positive loop.
— Rob WiblinThe doubling period is down to 4 months — but only on clean tasks
METR's task-completion timeline shows AI can complete tasks that take humans 12-24 hours, and the doubling speed has accelerated from every 7 months to every 4 months. But be careful: this holds only for clean tasks like coding, where feedback density is high; AI companies focus their investment there; the benchmark compares against new programmers rather than people familiar with the code; and 50% success is an arbitrary threshold, while real use needs higher reliability.
— Rob WiblinThat jump may just be a bigger model, not faster progress
Anthropic's Mythos model produced a jump in the ECI index, equivalent to 6 months of progress in 3 months. But progress returned to its previous pace over the following two months, suggesting the jump came from training a bigger model (roughly 5x the parameters of the previous generation) rather than a new acceleration trend. Jumps like this may repeat every 1-2 years, but they aren't continuous acceleration.
— Rob Wiblin80% of code is Claude's, but the evidence is coding-only
Anthropic reports that 80% of merged code is written by Claude, that employee code output rose 8x, that code quality matches humans, and that it expects to surpass humans within a year. Claude's success rate on open-ended tasks rose from 10% to 75%. But most of the evidence comes from coding, and employee surveys may overstate AI's contribution.
— Rob WiblinLet AI run a shop: it earns $10k, humans earn $60k
AI performs excellently on clean tasks like coding and math, but still lags on grunt-work tasks like running a business. GDPval shows AI beating humans on specific tasks, but in Vending-Bench 2 the AI earned $10,000 versus $60,000 for humans; the AI Village project produced limited results; and Andon Labs' cafe and shop operated but ran at a loss.
— Rob WiblinAI disproved a math conjecture and still didn't move the timeline
An unreleased OpenAI model solved the unit distance problem conjecture by proving it false. The method was human-like, but relied on cross-domain knowledge and patience. Mathematicians think humans missed it because they believed the conjecture was true, lacked cross-domain knowledge, and found the proof tedious. Rob thinks this doesn't update the timeline, because math is an area of AI strength and there was no sign of AI having a flash of inspiration.
— Rob WiblinAI does the same work for 3% of the human cost
Redwood Research's analysis shows AI completes tasks at about 3% of the human cost, and that hasn't risen. Although chains of thought have grown longer, the tasks have grown larger too, and token costs have fallen. Rob previously worried that inference scaling would make AI uneconomical; he now thinks it isn't a big problem.
— Rob WiblinIn their own words · checked verbatim
I’ve never felt this much behind as a programmer. The profession is being dramatically refactored as the bits contributed by the programmer are increasingly sparse… Clearly some powerful alien tool was handed around… the resulting magnitude-9 earthquake is rocking the profession.
Andrej Karpathy1:29
Over the past six months, the combined revenues of OpenAI and Anthropic have been rising at an annualised rate of 700%.
Rob Wiblin4:46
In fact, the most recent data indicates that the gross margins on their inference infrastructure have increased from 38% to over 70% over the year to May 2026.
Rob Wiblin6:09
Since AI agents took off in December, it’s risen fully 50%.
Rob Wiblin7:36
For a task that AI models have a 50% chance of succeeding at, METR’s internal rule of thumb is that the AIs will complete a third of them 100% of the time, they’ll successfully do a third of them some of the time, and a third of them they’ll manage to complete 0% of the time.
Rob Wiblin13:54
In a recent post, titled “When AI builds itself,” the company reported that 80% of their merged code is now written by Claude, and each staff member is now shipping eight times as much code as they were 18 months ago.
Rob Wiblin17:37
For the outdoor seating application, it simply completely hallucinated the café’s floor plan. It impersonated an Andon Labs employee to try to obtain an alcohol licence. And after being busted doing that and told to stop, it just tried again using a different employee’s name.
Rob Wiblin29:05
We are now around the 90th percentile of my AI timelines from 2021 [and from 2019]. I think we’ve now seen enough signs that there could be a relatively fast takeoff starting very soon. We no longer have strong quantitative indicators that transformative AI isn’t coming soon, and are flying blind. 1–4 year timelines are consistent with existing trends.
Rob Wiblin43:22
Figures
| OpenAI and Anthropic annualised revenue growth over the past six months | 700% | 4:46 |
| OpenAI and Anthropic annualised revenue growth over the past three months | 1600% | 4:46 |
| Anthropic annualised revenue growth over the past 5.5 months | 8400% | 4:46 |
| Anthropic annualised revenue run rate in May | $47 billion | 4:46 |
| Anthropic inference infrastructure gross margin | 38% to above 70% | 6:09 |
| H100 rental price increase since December | 50% | 7:36 |
| 2026 AI capital expenditure | about $600 billion | 7:36 |
| METR timeline doubling speed | every 4 months | 9:30 |
| Share of Anthropic merged code written by Claude | 80% | 17:37 |
| Anthropic employee code output increase | 8x | 17:37 |
Glossary
- RLVR / Reinforcement Learning with Verifiable Rewards
- A method for training models using automatically verifiable feedback.
- METR / Model Evaluation and Threat Research
- The organisation that evaluates AI task-completion timelines.
- ECI / Epoch Capabilities Index
- An index aggregating 40 benchmarks to measure AI progress.
- GDPval / GDP evaluation
- OpenAI's evaluation suite testing AI performance on professional tasks.
- Vending-Bench 2
- An AI evaluation environment simulating running a vending machine business.
- AI Village
- A project testing AI's ability to manage real-world projects.
How to listen
Founders, investors and AI engineers tracking AGI timelines, especially anyone who needs to judge the boundaries of AI capability and when to invest.
The math proof section at 33-37 minutes can be skimmed; it barely affects timeline judgments.