The grieving dog owner test

Last week a grieving dog owner closed the gap between "AI could theoretically do this" and "AI did this" in 60 days, for $3,000, with no biology degree. Last week Anthropic published a report explaining why corporations haven't.

The word "yet" is doing a lot of work in that second sentence.

The dog's name is Rosie. Most of her tumors have shrunk. She is back at the dog park chasing rabbits.

I am going to need a moment.

Okay. 𝗧𝗵𝗲 𝗽𝗵𝗮𝗿𝗺𝗮𝗰𝗲𝘂𝘁𝗶𝗰𝗮𝗹 𝗶𝗻𝗱𝘂𝘀𝘁𝗿𝘆, which has had decades, billions of dollars, and the combined intellectual might of some of the most credentialed humans alive, had not done this. One man who loved his dog and refused to accept the vet's prognosis did it in the time it takes most companies to complete an onboarding checklist.

𝗡𝗼𝘄 𝗯𝗮𝗰𝗸 𝘁𝗼 𝗔𝗻𝘁𝗵𝗿𝗼𝗽𝗶𝗰'𝘀 𝗿𝗲𝗽𝗼𝗿𝘁, which the internet immediately converted into either a "we're fine" headline or a "jobs apocalypse" headline, because nuance is a renewable resource we have collectively agreed not to use.

𝗧𝗵𝗲 𝗮𝗰𝘁𝘂𝗮𝗹 𝗳𝗶𝗻𝗱𝗶𝗻𝗴: AI can theoretically handle 94% of tasks in computer and math roles. It's currently doing 33% of them.

𝗧𝗵𝗲 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱 𝗿𝗲𝗮𝗱 on that 94% vs. 33% gap: the storm is loading.

𝗧𝗵𝗲 𝗿𝗲𝗮𝗱 𝗻𝗼𝗯𝗼𝗱𝘆 𝗵𝗮𝘀 𝗺𝗮𝗱𝗲 𝘆𝗲𝘁: that gap exists almost entirely because of institutional friction. Not capability. Not cost.

Somewhere between "theoretically possible" and "actually deployed" lives a 47-slide deck, three rounds of legal review, and a VP who needs to feel consulted.

𝗣𝗮𝘂𝗹 𝗖𝗼𝗻𝘆𝗻𝗴𝗵𝗮𝗺 𝗵𝗮𝗱 𝗻𝗼𝗻𝗲 𝗼𝗳 𝘁𝗵𝗼𝘀𝗲 𝗰𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝘁𝘀. He had a dying dog, $3,000, and a direct line to people who wanted to help. He didn't wait. He closed the gap himself.

The Anthropic report also 𝘁𝗲𝗹𝗹𝘀 𝘂𝘀 𝘄𝗵𝗼'𝘀 𝗺𝗼𝘀𝘁 𝗮𝘁 𝗿𝗶𝘀𝗸: older, educated, well-paid knowledge workers. Which is exactly the kind of person who has the skills and financial cushion to close that gap before their employer gets the steering committee together.

𝗧𝗵𝗲 𝟮𝟮-𝟮𝟱 𝘆𝗲𝗮𝗿 𝗼𝗹𝗱𝘀 who can't get hired into those same roles right now? The tools to not need that job cost $20 a month. This is either terrifying or extremely liberating, depending on which way you're facing.

Nobody is building a personalized mRNA vaccine for your career.

But the infrastructure to do the equivalent, in whatever field you're in, is already sitting there, mostly unused, while somewhere a committee debates the font on the pilot program proposal.

𝗥𝗼𝘀𝗶𝗲 𝗱𝗶𝗱𝗻'𝘁 𝘄𝗮𝗶𝘁 𝗳𝗼𝗿 𝘁𝗵𝗲 𝘀𝘆𝘀𝘁𝗲𝗺. Neither did her owner. The system, to its credit, has not weighed in on whether this was permitted.

So: in 3 years, do you think most knowledge workers will look more like the pharmaceutical industry in this story, or more like Paul Conyngham?