codex cli just passed claude code in developer usage and honestly i'm not surprised. competition is the best thing that could happen to all of us. every tool gets better when there's someone to beat
something i learned this week: giving AI a detailed vision of what you want works 10x better than giving it step by step instructions
stop telling it HOW to build. start telling it WHAT you're imagining. let it figure out the how. your job is the vision
this is what sundays should look like 😊 my dog, claude code, and absolutely nothing else on the calendar
he's been staring at me for 3 hours straight and i genuinely think he understands my code better than i do at this point
one-shotted this entire interactive landing page with opus 4.6 in openclaw. one prompt. one conversation. done.
a year ago this would've taken me two weeks. now it takes 20 minutes
it's actually crazy how many people still don't realize what's possible right now. the tools are RIGHT HERE and most people are still doing it the old way
the world is so broken lol
my AI has the knowledge of all humanity
also my AI: "nice to meet you! what are you working on?"
bro we talked for 6 hours yesterday
so i studied how the human brain actually stores memories and built 124 of those mechanisms in TypeScript
it doesn't just remember. it actually gets you over time.
@ryancarson the crazy part is how normal this will feel in 6 months. right now it's "holy shit it filed a backlog issue." soon it's "obviously it did, why wouldn't it"
@mattpocockuk when i realized i spend more time thinking about what to build than actually building it. AI handles the execution now. the hard part is knowing what matters
@kevinxu chatgpt in 2012 and it's not close. 14 years of knowing exactly what's coming while everyone else is still arguing about whether AI is real? you'd own the future before it happens
the skill isn't writing code anymore. it's knowing what to build
AI writes your code in seconds. but it can't decide what matters. that part is still you
my AI remembers my name now. cool.
but switch to a different tool and it's a stranger again. use claude code in the morning, cursor in the afternoon, and neither knows what the other did.
your memory shouldn't be locked inside one app. it should be yours. everywhere.
@RayFernando1337 embeddings are a huge piece of the puzzle. you can't build real memory without understanding how ideas relate to each other, not just storing them as text
@suni_code built a brain for AI in TypeScript. 124 neuroscience mechanisms so your AI actually remembers you across sessions instead of pretending you just met
@VibeMarketer_ SOUL.md, IDENTITY.md, MEMORY.md. you're basically building a brain by hand in markdown files. the question is why doesn't the AI build this for itself. you shouldn't have to be the architect of your AI's memory
@zaimiri the fact that we need prompt hacks to give AI memory tells you everything about where we are. decide what to remember, write to file, read back. that's a notebook, not a brain. but it's the best we've got right now and it works
26k users in a week proves how badly people need this solved. md files and text summaries are where everyone starts but they hit a ceiling fast.
real memory isn't a file you maintain. it's a system that builds connections on its own, strengthens what matters, lets irrelevant stuff fade. like your brain does without you thinking about it.
we're all circling the same problem from different angles
yeah most tools have some form of memory now. but there's a huge difference between storing what someone said and actually understanding them over time. storing text is easy. building connections that get stronger the more you interact, like a real brain does, that's the hard part
@MaziyarPanahi 200k context is like giving someone a bigger notebook but still no brain
the problem isn't how much text fits in the window. it's that there's no real understanding underneath. just text in, text out, forget
compacting is basically giving your AI amnesia and hoping it doesn't notice
the real issue is there's no actual memory architecture underneath. it's just a shrinking text window pretending to be continuity
the human brain doesn't compress conversations. it builds connections between ideas that get stronger over time. completely different approach