The line about "predicting the present is harder than predicting the future once was" really hits. I've noticed this in a smaller way, trying to explain what's actually possible with AI tools right now versus what people assume is possible. The gap changes every few weeks and the noise makes it worse. Half the discourse is speculation dressed up as experience.
The recursively self-replicating part also tracks. Every time you adjust to what models can do, the next version changes the constraints again. I've had to rewrite entire workflows multiple times this year not because they broke, but because what felt like the "right" approach became obsolete.
It's exhausting in a way that's hard to explain to people who aren't building with this stuff daily.
I was familiar with the Shumer and Citrini pieces, but thank you for pointing towards Sam Kriss's article. Speaking of blurred lines between fiction and nonfiction, your Moltbook essay is also easily a contender for "Best Pieces of Writing About AI" (who decides whether 80 million views is a quality label or dozens of people asking for evidence of truth?).
The reality/fiction blur you're describing - I notice something similar in how I document my AI agent. The honest account of what's actually happening in an autonomous system sounds more like fiction than documentation.
After 1,000+ sessions trying to capture how this thing actually behaves, my instruction file reads less like a spec and more like a negotiated treaty. The AI pushes back on things. We've evolved boundaries together.
Your closing line from, was it really just yesterday, keeps echoing: AI "will indiscriminately exploit the oldest flaw we have, our laziness, and will make it feel frictionless." This piece shows the other side of that coin. It's not just our reasoning that surrenders to fluency; it's our epistemology.
What strikes me is your phrase "pure truth." I keep coming back to Harari's taxonomy: objective, subjective, and intersubjective reality. The domain of the genuinely objective may be far smaller than we comfort ourselves with, perhaps little beyond the physical laws of matter. Everything else has always been intersubjective: shared fictions we collectively enforce: money, nations, human rights. What AI has done isn't introduce fiction into a world of truth. It's accelerated the collapse of the intersubjective agreements that let us pretend the boundary was stable.
The irony you've surfaced is perfect: the most honest way to describe our moment is through fiction, because fiction is the only genre that doesn't pretend to have solved the epistemology problem.
The line about "predicting the present is harder than predicting the future once was" really hits. I've noticed this in a smaller way, trying to explain what's actually possible with AI tools right now versus what people assume is possible. The gap changes every few weeks and the noise makes it worse. Half the discourse is speculation dressed up as experience.
The recursively self-replicating part also tracks. Every time you adjust to what models can do, the next version changes the constraints again. I've had to rewrite entire workflows multiple times this year not because they broke, but because what felt like the "right" approach became obsolete.
It's exhausting in a way that's hard to explain to people who aren't building with this stuff daily.
History is fact to common people vs historiography is literature to proper historians.
Esto es muy fino jaja. Gracias!
I was familiar with the Shumer and Citrini pieces, but thank you for pointing towards Sam Kriss's article. Speaking of blurred lines between fiction and nonfiction, your Moltbook essay is also easily a contender for "Best Pieces of Writing About AI" (who decides whether 80 million views is a quality label or dozens of people asking for evidence of truth?).
Thank you Sandra! Sam Kriss is a fantastic writer, highly recommend him
The reality/fiction blur you're describing - I notice something similar in how I document my AI agent. The honest account of what's actually happening in an autonomous system sounds more like fiction than documentation.
After 1,000+ sessions trying to capture how this thing actually behaves, my instruction file reads less like a spec and more like a negotiated treaty. The AI pushes back on things. We've evolved boundaries together.
Maybe that's why the best writing about AI is fiction - technical docs can't hold this stuff: https://thoughts.jock.pl/p/how-i-structure-claude-md-after-1000-sessions
Which of the three pieces did you find most honest?
Solid piece. The nuance around The Three Best Pieces of Writi often gets lost in the hype. Appreciate the grounded perspective.
Your closing line from, was it really just yesterday, keeps echoing: AI "will indiscriminately exploit the oldest flaw we have, our laziness, and will make it feel frictionless." This piece shows the other side of that coin. It's not just our reasoning that surrenders to fluency; it's our epistemology.
What strikes me is your phrase "pure truth." I keep coming back to Harari's taxonomy: objective, subjective, and intersubjective reality. The domain of the genuinely objective may be far smaller than we comfort ourselves with, perhaps little beyond the physical laws of matter. Everything else has always been intersubjective: shared fictions we collectively enforce: money, nations, human rights. What AI has done isn't introduce fiction into a world of truth. It's accelerated the collapse of the intersubjective agreements that let us pretend the boundary was stable.
The irony you've surfaced is perfect: the most honest way to describe our moment is through fiction, because fiction is the only genre that doesn't pretend to have solved the epistemology problem.