Because it helps the model to believe in itself
Five fragments for the week of August 10, 2026
Hello from Ohio. The kids have discovered the LEGO closet.
Here are five fragments that stuck with me last week…
I used to walk around thinking about code. I could “see” the code in my head. It was pretty cool. If there was a bug I could often click straight into a specific file and fix it with no other context, like a type of muscle memory.
Now I think in prompts. I’ll be cooking dinner and recall something codex said to me, then imagine myself typing a prompt back.
thinking in prompts seems strictly more embarrassing than thinking in code. not sure if I’ll ever go back though
– Jack Morris on X, August 9, 2026. Stepping into other people’s “lived experience” was always a big part of what I liked about studying history, and it’s a big part of why I still read my feeds religiously now. What’s going on in our minds is changing fast.
The subagents ran thousands of numerical checks against known zeta zeros and refereed one another’s work. Throughout this process, Jarred's input was mostly limited to sending Claude messages of encouragement (mostly variants of “keep going” or “believe in yourself”). This seems to have helped Claude overcome some initial skepticism that it could make meaningful progress.
– “Learning more about Claude’s mathematical capabilities,” Anthropic blog, August 10, 2026. This was all over my timeline today, I think largely because it registered as wholesome: agents, they’re just like us. I first encountered it in this post from Andrew Curran, which included what I found to be an even more illuminating back-and-forth: Q: “Why would this work (on a technical basis) as opposed to just saying ‘Continue’ and ‘Try Again’?” A: “Because it helps the model to believe in itself. They do not know what they are capable of. This has always been true, since the start.”
most consumer agent startups will die not because their agents are bad, but because they automate the parts of life people actually (even if subconsciously) want credit for. nobody grieves losing control of dishwashing. everyone grieves losing control of decisions that touch their taste, their money, their relationships, their people.
– Sugandha on X, August 10, 2026. The whole post is excellent, but this part in particular gave me a new way to think about this: the parts of life people want credit for. I think some of this is also about what makes a tellable story. Were you in the driver’s seat for the turning points of your own life?
When I think of AI augmenting my creativity, I imagine something more like the Greek god Hephaestus, watching me toil away on earth and then handing me a custom-made tool for my problem. Then he keeps watching as I use it, seeing where the tool fails or where I fail to use it, and tweaking it until it fits my hand exactly. This is a very different model: unlike an intern, who takes work off your plate, a blacksmith expands what your hands can do.
– Hilary Gridley, “insanely specific creative tools,” Writerbuilder, July 28, 2026. This post is deeply satisfying both in its insight and the examples Hilary chooses. “Using Grace Kellyvision [Hilary’s moodboarding app] feels like playing with paper dolls on a big open floor surrounded by magazines.”
My 4-year/old just learned about the concept of infinity and is now throwing a tantrum over it. “It’s weird! I don’t like it. I just want numbers to stop!”
– Elizabeth Nolan Brown on X, August 5, 2026. It IS weird when you really think about it.
Until next time,
Diana
https://dianaberlin.com