Living at the hinge of history - by Henry Stanley
Aliveness Studies
SubscribeSign in
Living at the hinge of history<br>It's an insane time to be alive<br>Henry Stanley<br>Aug 22, 2026
Share
Wanderer above the Sea of Fog, Caspar David Friedrich (1818)<br>As a software engineer, I’m at the pointy end of watching AI automate me out of a job (perhaps just behind mathematicians, who are seeing theorems which have been unsolved for hundreds of years get cracked by someone with no math knowledge telling Claude “just try harder, believe in yourself”).<br>I recently started a new role after being unemployed for a few months. In that time, my job has entirely changed. My whole day is now rallying agents, giving them work (often in the form of “do this ticket ”) and monitoring them. Even the monitoring can be automated – I ask Claude to do a code review of its own work, then push its code, wait for our other code review service to review it once more, and implement the feedback. And Claude can go onto our staging site and QA its own work, click-by-click.<br>If it didn’t already feel so mundane it would be literally unbelievable. What these machines can do is unimaginable. It feels like anything is possible, but it also feels like I’m an idiot giving work to an enormously more competent and vigilant colleague.<br>If you aren’t in software, it’s hard to convey how powerful these systems are. They don’t just write code – they act, autonomously, in the world. They can plan whole new features, implement them, fix the bugs, get them deployed, and then open a browser and run a full end-to-end test. They can reconfigure your computer. They can do online shopping for you, suggest things you might like to buy and order them to your house (and you can say “get something for yourself” and it’ll buy a cat figurine). They have taste, they have opinions, they’re fun to work with.<br>Soon they’ll be able to work autonomously for hours, or days. Already I’m cutting myself out of the loop. The agents work for me like employees, not tools. They work all the time and put in superhuman amounts of effort. The prophecy of “a country of geniuses in a datacenter” is coming true. It’s insane to have a front-row seat to the advent of a technology that’s closer to the invention of the transistor than the internet. We are taking sand and making it think1.<br>I don’t really know how long I’m going to be useful for.
With it comes a kind of ennui.
I want to live the slow life. I moved to Berlin to have my hot girl summer (maybe several), but the self-imposed pressure to work is always there. I work for a nonprofit, so more work = more money donated = more lives saved, more animals protected from harm, the future defended against catastrophe. Previously the startup cost was high; you had to re-load all that code into your short-term memory and it was a slog. Now I can just run claude agents and start throwing them some work. There’s always something to be done, some fruit that’s now descended to be low-hanging thanks to the latest model release, or bonus tokens to use.<br>Is now the time to eke out as much productivity as I can, before my skillset is no longer useful? Or is that a hiding to nothing, given in six months the agents will work 10x as fast, so I might as well just wait for that to happen and enjoy myself in the meantime. Junior devs I know are already unable to get jobs, and maybe never will.<br>The same with donations – an unfathomably large glut of funding is going to appear when early Anthropic employees’ equity is unlocked by their pending IPO. An amount so large it’s equal to multiple Bill Gates foundations, a dozen GiveWells. The transfer to nonprofits will be enormous – maybe my time is better spent shoring them up and getting them ready to absorb that cash.<br>Or maybe now is the time to donate everything, a substantial chunk of my net worth, to front-load the work we can do before that money arrives. Probably that looks like lobbying to get the AI labs to slow down, not allow rogue agents to escape and hack into secure systems again and again and again, stop building systems they are unable to control, and stop encouraging China to catch up to the advancing capabilities baselin. (Us effective altruists have much to answer for.)
One bit of weirdness is the way Claude in Slack came full circle. The very first private preview in Claude was an Anthropic Slack channel. You sent a message, and Claude pre-1.0 would reply in a thread. This was before ChatGPT. And you could see what everyone else was asking, and how they were getting to grips with the bizarreness of the thing, along with all the confident hallucinations that plagued models until relatively recently.<br>Now, Claude (as Claude Tag) works for my team autonomously in Slack. It doesn’t even need to be mentioned by name – it’s scanning all the messages quietly in the background, offering its thoughts or context where it can, otherwise staying out of it. It feels like a superhuman human colleague.<br>When I...