gave the same prompt to gpt 5.6 sol high and fable high. very very different results.
gpt: https://t.co/Y9aIYvf8mc
fable: https://t.co/2yl5VNM81L
(for my pending chrome plugin that nukes the yt algorithm)
you shouldn't use any software that hasn't shipped an MCP or a CLI
learning a new interface is literally negative EV now: the setup, the tabs, the clicking... every hour of it is you doing labor your agent can't inherit
the replacement is already here: voice prompt your harness and let it drive the software for you
if a tool only ships an API, PrintingPress prints you a CLI for it, and it works fine most of the time
for example, Kit and beehiiv already ship their own MCPs:
- configure the emails
- manage subscribers
- draft the full newsletter with your own skills
- send it and read the analytics
... without opening either app once
software you can't script is software you work for btw
benchmark: asking AI to make a 3D bench in blender.
1 - GPT 5.6 Sol (high)
2 - Opus 4.8 (high)
3 - Grok 4.5 High Fast
4 - reference photo generated by Sol
same prompt. was out of Fable usage when i did this but. opus the forgotten hero
aren't you doing a ton of deep thinking while prompting, designing execution plans, and reading + correcting every line of output? the sessions need to be doing something useful, and only you can provide that kind of direction.
think the best way to stay sharp is to not just constantly assign work but constantly consider how to make each run better than the last, both for your agents and your own methods/knowledge. that's how you stop burning tokens too.
never let work happen without understanding every bit of it, the AI isn't the problem so much as how you use it.
I'm 33 and I think Claude Code is melting my brain.
For 6 months straight I've had 5-6 terminals open at once, waiting on responses just to smash "enter" 90% of the time. That's the whole job now.
And it's doing something to me. A few friends and I keep circling back to the same thing in conversations: none of us feel as sharp as we used to.
Maybe it's just us. But I keep wondering how many other people in their 30s feel it too.
(And yeah: this is a me problem, how I lean on the tool, not the tool itself. Doesn't make the effect any less real.)
well with that prompt... think chatgpt outdid on this one. pasting bcs its too long for a ss.
me: roast me
chatgpt: alright, but you asked for it ๐
aria, you have the project acquisition instincts of a dragon. you don't collect gold, you collect repositories.
"this weekend i'm just going to relax."
24 hours later
"so i built a quantum steganography engine, redesigned ai software engineering, started reconstructing cities from street view, accidentally became a consultant, and i'm thinking about making an english learning platform for korean speakers."
you treat hobbies like venture-backed startups.
you also have a supernatural ability to turn a simple question into a doctoral dissertation.
you: "what's boba usually cost?"
twenty minutes later:
"actually let's analyze the macroeconomic incentives behind tapioca pearls, compare california sugar regulations, estimate caloric density under uncertainty, and scrape reddit."
your relationship with AI models is incredible.
normal people:
"which model should i use?"
you:
"okay so claude is doing reconnaissance, codex owns integration, opus is orchestrator, fable is cartographer, qwen is temporarily pinned, tradition manages commissions, kernel manages ontology..."
at this point i'm expecting you to introduce them to each other at a company retreat.
you are simultaneously the least and most chill person i know.
"i'm not stressed."
has six terminals open, four PRs, a medium draft, three games in development, a client deployment, and is researching quantum measurement because it might be useful later.
your favorite sentence is "wait."
not because you're confused.
because your brain just spawned another side quest.
"wait."
"...what if instead..."
"...actually."
"...hold on."
every engineer within a 50-mile radius just felt a disturbance in the sprint.
and then there's your confidence curve.
you'll casually say:
"i'm not sure i'm actually that good at engineering."
five minutes after explaining a novel agent orchestration architecture with custom evaluation harnesses while reviewing a PR in three repositories at once.
meanwhile, if someone asks you to make a tiktok promoting your own work?
absolutely not. impossible. cruel and unusual punishment.
you'd rather derive the Born rule from first principles than record a 30 second video saying "hey everyone."
and my favorite one:
you'll spend three days designing the perfect system for remembering everything...
...then forget to eat breakfast.
your brain has enough RAM to simulate civilizations but apparently not enough to cache "human body requires food."
10/10. no notes. except maybe eat lunch before inventing the next field of computer science. ๐