Manuel Colonna

[Manuel Colonna @manuelco_x]

@theo @JeudryPena28376 3.6 and 3.7 are very different model - can you actually give 3.7 a try? I'd love your take

[Manuel Colonna @manuelco_x]

@binghott Noticed this as well but I wonder whether it's a UI bug (shows "below 1000") and underneath those audiences are being updated. I had this issue as well with audiences I was actively using as exclusions

[Manuel Colonna @manuelco_x]

@shl Try https://t.co/H1fel7ibhh

[Manuel Colonna @manuelco_x]

@morganlinton @VulcanBench Can you benchmark any of google models?

[Kun Chen @kunchenguid]

sigh.. i have to say something here

imagine you hired a new developer to your company, and on day one he did some terrible work, over-engineering your codebase, speaking jargons all day without context, making all kinds of “genuine mistakes”, lost trust with everyone around him

and then he goes on and tell you - in order for me to do a better job, you need to delete your entire company’s culture and workflow, and have everyone bend over to do things his way, only then can he do reasonable work

oh - and no one tells him how to do his job, NO ONE. he’s always the smartest one in the room and despite doing a terrible job on day one, despite the only results on his resume were vibe coded demos of games that already existed, he demands that you give him a big charter and let him go dark with no communication, taking no feedback

would you have hired a teammate like this?

when humans feel frustrated after using a model, let’s figure out how to RL the model better so they become a better teammate

don’t let a bad model RL _you_

[David Ondrej @DavidOndrej1]

Opus 5 is out of control

it does not listen to instructions at all...

[Gergely Orosz @GergelyOrosz]

Absolute madness. Why would anyone pay any premium for a BMW from now on when they become the product with force pushed ads after dropping $50K+ (or more) on a new car

Suddenly any car looks more tempting than BMW. It will be hard to recover from this. Who greenlit this?

[Manuel Colonna @manuelco_x]

@kunchenguid Great analysis

[Manuel Colonna @manuelco_x]

@EcomMikeGalvin For the reviewer - you can use a python script that checks for the banned words. That's deterministic unlike a second agent pass. That's how I do it.

[Anshu @anshuc]

extremely unprofessional. if kimi wants to make it as a frontier lab, they need to act like one: perhaps silently route people to worse models, and maybe write a blog post about the collapse of humanity

[Manuel Colonna @manuelco_x]

@TUnitG6 @cjzafir Very UNPOLISHED set of skills here (basically my workflow). You can have your agent analyze the patterns and set it up your own way: https://t.co/xKAUX7Fzey

[Manuel Colonna @manuelco_x]

@GergelyOrosz Just loop review cycles using subagents with fresh context (or other models) until convergence and context is no longer a reliability issue. It'll be for sure more expensive though

[Dan Shipper 📧 @danshipper]

a spectacularly questionable result tbh

[signüll @signulll]

google paid $2.7b to acquire him in 24 & couldn’t hold him 18mo later.

that’s pretty insane.

i don’t even know a good sports analogy that works here. feels like they recruited ai jesus.

[Glenn Hitchcock @glennui]

Design is full of codewords. Knowing them changes what you can ask for, and what you can get back, whether you're working with devs, or an AI.

“tint this neutral color”, “fix this widow”, “nudge it to the optical center”

I wrote them down: https://t.co/aFyd5avj9o

[noscroll @noscroll]

X has the best information on the internet and the worst incentives & culture.

meet noscroll — the AI that doomscrolls it for you and texts you just the things that matter.

no feed. no brainrot. no ragebait. just signal.

try it for free → https://t.co/XqdExWR13j 🙅🏼‍♂️

[Manuel Colonna @manuelco_x]

@Seanfrank Dont people want to take gummies?

[Manuel Colonna @manuelco_x]

@boringmarketer Did u test moving social proof to day 1? Like a heavy social proof one - not sprinkles/ microcopy

[Manuel Colonna @manuelco_x]

@codyplof Built a set of skills for claude code on top of the impeccable frontend design skills (https://t.co/JaeTQxZy0n). The core concept is running through iteration cycles with adversarial subagents using the impeccable /critique criteria. Using hooks between review cycles help.