You Retweeted
RT @DavidDuvenaud: @geoffreyirving We tried that! The vintage models can just barely start to do simple things with Python, purely from in…
/tweets
My Twitter/X archive going back to 2014. 5,809 tweets across 13 years.
//
Replying to @status_effects
or as @adagradschool would say... [cybernetic] vs [skeumorphic] applications of AI! t.co/s94OJq5nOu
Replying to @rcwhiteley
@rcwhiteley true, but i'd argue that they're more similar than most people think
Replying to @sofvanh
@sofvanh esp amongst builders/creatives/etc cuz it's interwoven with their identities (mine too)
i wish there was a way to replay codex's computer use traces in fast forward mode. it's v hypnotic to see the cursor moving about, clicking, and doing things.
my hot take is that "taste is a uniquely human thing" is cope.
one simple example: i've noticed that friends of mine who are not designers at all (for example, backend / infra people) were making projects with ai 2 years ago that looked obviously sloppy.
now the same people are shipping things that look… pretty polished. probably good enough that a junior designer could pass it off as their own work.
and when i ask them what changed -- did you learn some design fundamentals? did your 'visual taste' improve? are you better at noticing bad spacing / typography / color hues?
most of the time the answer is just: the tools got better. the models got better.
i can only judge this for visual design / frontend because that’s where i feel like i have some taste. but i’m pretty sure the same thing is happening in copywriting, backend architecture, research ideas, product taste, etc.
yes, prompting matters. yes, having a good taste yourself still matters. but the median 'taste' of an ai model in 2023 and the median taste of an ai model in 2026 are very different.
there is a waterline of taste and it is rising very fast.
sloppification of the internet was a temporary phase.
You Retweeted
RT @heyglif: Excited to officially announce Glif V2 and our $17.5M seed round!
it’s wild what @heyglif agent can do in one conversation -- bring your best prompts! t.co/jzL8xHCaU6
You Retweeted
RT @0x_kitsune: To everyone whining: this is cyperpunk dystopia as fuck and therefore cool. If this was Neuromancer there would be a whole…
Replying to @WeirdoWizard
@WeirdoWizard i think this was the midjourney video model when it first came out!
Replying to @WeirdoWizard
@WeirdoWizard we have come a far way... t.co/QTUvwApcsl
mcp servers that shouldn't exist -- took a while, but very much worth it! t.co/T5fmHXvYgF
i think opus-4.7 is a genius model -- and when the dust settles, it will be very obvious.
it's just the case of people evaluating the results on old harness optimized for older models.
is it possible for @openclaw to use codex's computer-use? t.co/DhjXiJDEsQ
Replying to @alexalbert__
@alexalbert__ but... does prompt caching carry over when advisor call is made?
Replying to @dhvanil
made with @heyglif in <30 min, and a handful of prompts!
the real AI eval isn't benchmarks
it's giving one identical twin to Claude and the other to ChatGPT and watching what happens to their lives
twin studies, but for LLMs.
watch what happens... t.co/Q6QZv0aXmI
You Retweeted
RT @baroquespiral: “the technocapital superintelligence stans Mark Fisher” might be the biggest Nick Land L of the millennium so far
Replying to @usgraphics
@usgraphics how does it emerge from the bottom up? or like... are there good examples of bottom up emergence of top grade aesthetics?
i've always assumed it needs to be top down 🤔
is this a convoluted plan by meta just to train their models on claude's agentic traces and outputs? 🤔 t.co/jm4m9j08jT
ai models are named like hurricanes now. the scarier the name, the more you’re supposed to pay attention
it always surprises me how good codex is at git surgery
sometimes i give it a large exploratory branch, and ask it to make it production ready. and it extracts the vibecoded slop to manageable PRs, keeps the diffs reviewable, writes nice descriptions, handles merge conflicts gracefully, preserves the original intent, etc!
You Retweeted
RT @fabianstelzer: super stoked to be co-sponsoring @levelsio's #vibejam with @heyglif !
It's never been easier or more fun to build GAME…
when i do multi agent shenanigans, and pipe some workflows to a new agent/team -- i love to tell the original agent(s) to "stand by"
i know it doesn't do anything, but it always gives me a nice feeling when i do that!
You Retweeted
anecdotal experience, but seems like today's models are much worse at few shot prompting -- than even gpt-3 era models!
You Retweeted
You Retweeted
RT @bandiaditya: I’m thrilled to announce we’ve raised $44M to build a new home for product design. Meet @noondesign.
No workflow is more…
Replying to @phileisn
@phileisn sometimes it throws very weird curveballs at me.
"i pushed the commit. if you want, i can change the name of the branch to (something completely random) now" 😅
codex really likes to say "let me XYZ, so I can answer concretely instead of hand-waving."
Replying to @LukeW
@LukeW kanbans, calendars, inboxes, dashboards are skeuomorphic by nature. my hunch is that we need more cybernetic interfaces
t.co/Z79db4BtZy