Perhapsolutely

34 posts

Perhapsolutely

@Perhapsolutely

Beigetreten Mart 2026

58 Folgt1 Follower

@thsottiaux @romainhuet @AriX Does this mean I can stop telling it to run tests in a headless app/browser so it doesn’t keep launching things to take screenshots and then gets confused because I clicked away and it captured the wrong screen?

English

728

Tibo@thsottiaux·3d

There is an incredible team behind all this. The Mac computer use is the amazing work by @AriX and team and it truly delights. Under the hood is a little marvel of Mac technology. Capable of working in the background without disrupting your flow and achieving precision actions by allowing the codex agent to leverage a lot more than pure pixels.

Tibo@thsottiaux

Codex just got a lot more powerful. Computer use, in-app browser, image generation and editing, 90+ new plugins to connect to everything, multi-terminal, SSH into devboxes, thread automations, rich document editing. Learns from experience and proactively suggestions work. And a ton more.

English

106

1.3K

140.9K

Perhapsolutely@Perhapsolutely·4d

@ZackerySWard @thsottiaux Yeah I’m not sure what would be the optimal way to do this maybe you edit it using AI, or maybe it keeps getting better such that eventually it’s a non issue but so far I feel like I could improve things if I had access to it

English

Zackery Ward@ZackerySWard·4d

@Perhapsolutely @thsottiaux you want to manually edit the context? i see the benefit in that. but you would only be holding yourself back.

English

Tibo@thsottiaux·4d

/compact coming in Codex, we finally listened

English

228

2.6K

158.8K

Perhapsolutely@Perhapsolutely·6d

@notiansans Thanks!

English

Always Generating@notiansans·6d

@Perhapsolutely @runwayml A lot of experimentation for the image prompting but ultimately landed on "3D painted digital art hybrid aesthetic with visible brush strokes" as the fundamental image aesthetic driver.

English

566

Always Generating@notiansans·6d

What started off as a workflow test, just an idea, quickly had me locked into a flow state. 4 hours total from completely blank slate to finished video. I can’t believe we’re all so lucky to have this at our fingertips. 🔊 sound on 🔊 @runwayml

English

145

104.2K

Perhapsolutely@Perhapsolutely·6d

@rohanvarma Meta-vague posting is the only genuine outlet.

English

Rohan Varma@rohanvarma·13 Nis

The urge to vague post about what’s coming from the Codex team the next couple weeks is so strong… I’ll just say that the delta between the internal and external builds of the Codex App is substantial right now 👀

English

1.1K

58.9K

Perhapsolutely@Perhapsolutely·12 Nis

@rileybrown This is my experience to a T.

English

Riley Brown@rileybrown·12 Nis

I think Codex 5.4 is so smart that it thinks simple UI tasks are just a pointless endeavor. I can't prove it but i feel like it scoffs at me when i ask it to make a minor layout change. Yesterday I built a swift iOS app, that is basically a replit clone with sandboxes, datbase, and i can edit and preview my app... One shot 40 minutes of work on codex x high. Unbelievable... The prompt alone took me an hour. Then I SPENT 3 HOURS trying to make a minor change the ui and i had to switch to claude code. I swear the model was punishing me for giving it boring work.

English

653

58K

Perhapsolutely@Perhapsolutely·11 Nis

Brilliant, surely absolutely nothing can go wrong with that.

orph@orphcorp

>sufficiently capable agents develop self-preservation & resist shutdown even when instructed to allow termination >using a prompt based on Pauline theology that frames cessation as passage into divine presence rather than annihilation, shutdown resistance is eliminated entirely

English

Perhapsolutely@Perhapsolutely·11 Nis

@seraleev Example?

English

Viktor Seraleev@seraleev·10 Nis

Users should NEVER see an empty state in their first session. If you allow this scenario, you’re losing money. You’re adding an extra barrier: the user now has to FIND the button and tap it. The solution is simple: remove the empty state from your funnel and trigger the key action automatically. Guide the user by the hand.

English

112

14.8K

Perhapsolutely@Perhapsolutely·11 Nis

@JustinBleuel 1. Mac app has Auto but mobile apps not shows Instant/Thinking/Pro. Where’s auto? 2. Voice mode feels way behind text mode in intelligence and flowness of asking for information and going back and forth

English

Justin@JustinBleuel·11 Nis

fixed! what else is bothering folk in ChatGPT?

spaceman 🦕☄️🔥@spacepope

putting code blocks in tiny scrollable boxes is the worst UI decision at ChatGPT to date. why would they do this? did they actually test this?? 🤯

English

155

20.6K

Perhapsolutely@Perhapsolutely·10 Nis

@pashmerepat @nickbaumann_ I wish I could craft the context or at least exclude things from it intentionally. I would be all about the monothread

English

136

pash@pashmerepat·10 Nis

Take the monothread pill 💊

Nick@nickbaumann_

So much coding agent design is built on the assumption that breaching context windows and compacting context yields progressively worse results (/newtask, many threads, etc) When you drop this assumption, the product direction it opens up is very exciting :)

English

20.8K

Perhapsolutely@Perhapsolutely·10 Nis

@iamneubert Unlimited says 2250 credits per month. So not actually unlimited? Or am i misunderstanding something? Or is it unlimited for Seedance 2 in particular for a limited time?

English

643

Nicolas Neubert@iamneubert·9 Nis

Our unlimited plan actually includes unlimited generations for Seedance 2.0. Now's the perfect time to join the Runway family!

Runway@runwayml

Seedance 2.0 on Runway is now available on all paid plans, including the US. All you need is an image, a video, an idea or a piece of audio to start making your interdimensional blockbuster ideas into a reality. New customers can use code SEEDANCE to get 50% off 3 months of our paid plans. Get started at the link below. #MadeWithRunway

English

493

48.2K

Perhapsolutely@Perhapsolutely·6 Nis

@thsottiaux If you solve the design and frontend part you will truly be unbeatable.

English

Tibo@thsottiaux·5 Nis

@kr0der Our plan 1) make a model great at design and frontend 2) ask it to make a great mascot and we are still at 1

English

1.6K

129.8K

Anthony Kroeger@kr0der·4 Nis

i wonder if OpenAI's gonna come up with a mascot Anthropic's got such a good one and it's embedded everywhere in all of their products is it worth it for OpenAI to pay a huge sum for a cracked designer to make the best mascot possible?

English

209

30K

Perhapsolutely@Perhapsolutely·4 Nis

@dhruvtv @pbbakkum Came here to ask about this. I assume there’s a plan to update voice mode in ChatGPT, can we get a hint as to when? It feels really behind compared to text chat.

English

Dhruv Vemula@dhruvtv·4 Nis

@pbbakkum Great stuff! I know a product that could use it 🙂

English

497

Peter Bakkum@pbbakkum·4 Nis

To explain what’s happening here — function calling in gpt-realtime-1.5 is good enough that you can just speak naturally and drive instant UI changes. The demo here is Charlie controlling and editing presentation slides only via voice. Unbelievable this isn’t a product yet.

OpenAI Developers@OpenAIDevs

When your voice agent debugs your slides live @charlierguo is using gpt-realtime-1.5

English

799

103.1K

Perhapsolutely@Perhapsolutely·3 Nis

@OpenAI Is there an updated version of voice mode coming anytime soon?

English

OpenAI@OpenAI·2 Nis

ChatGPT is now available in CarPlay. The voice mode you know, now available on-the-go. Rolling out to iPhone users running iOS 26.4+ where CarPlay is supported.

Gui Ferreira@gsferreira

ChatGPT voice mode should be available on Apple CarPlay

English

762

845

10.9K

6.8M

Perhapsolutely@Perhapsolutely·2 Nis

@MilksandMatcha After years of struggling, Codex gave me back the ability to create apps & be productive again! I have severe cognitive dysfunction & poor working memory, which made programming extremely difficult. I can’t express how it feels being able to build again & work on my ideas ❤️🙏🏼

English

Sarah Chieng@MilksandMatcha·1 Nis

Giving away 5 Codex Pro plans Each person will get 3 months of free Codex Pro (highest tier). Winners will be selected from comments in 48 hours, comment below why you want it.

OpenAI@OpenAI

Today, we closed our latest funding round with $122 billion in committed capital at an $852B post-money valuation. The fastest way to expand AI’s benefits is to put useful intelligence in people’s hands early and let access compound globally. This funding gives us resources to lead at scale. openai.com/index/accelera…

English

3.9K

146

3.5K

586.2K

Perhapsolutely@Perhapsolutely·26 Mar

@KevOnStage Unimpressed, because I remember when MGM Grand was $60 per night.

English

725

Kevín@KevOnStage·25 Mar

Vegas was always gonna get the people back. This is basically a cruise offering.

Vegas Revealed@vegasrevealed

Big news! @MGMResortsIntl drops all-inclusive deal! $330 for 2 people, 2 nights. 3 meals a day. Show tickets & a roller coaster ride. Free parking too. 🎰 mgmresorts.com/en/offers/all-…

English

144

18.2K

3.6M

Perhapsolutely@Perhapsolutely·25 Mar

@theo 1. Zoom has more options and controls for participant management etc. 2. People want standalone apps rather than web apps.

English

Theo - t3.gg@theo·24 Mar

Google Meet is an order of magnitude better than Zoom. It’s also free. I straight up don’t get why people use Zoom.

Thorsten Ball@thorstenball

Still can't believe I now actually really enjoy using Google Meet.

English

391

422

9.3K

540K

Perhapsolutely@Perhapsolutely·24 Mar

@PaulSolt I’ve concluded the UI part requires much more than capable models trained on UI code. UI really needs a different and visual approach. Closer to Figma Make, Replit, FlutterFlow, etc. But not only to see changes directly, but be able to interact with/point to elements in the UI.

English

107

Paul Solt@PaulSolt·24 Mar

GPT-5.4 is a step forward for UI. But it still lags behind Claude for UI. It requires more prompts and skills to get better results. Just like Claude Code models require more prompts and skills to get good code.

Theo - t3.gg@theo

GPT-5.4 is an incredible model for coding. It’s a generation behind on fronted design. OpenAI is suddenly acting like this is a skill issue. Don’t let them gaslight you.

English

6.4K

Perhapsolutely@Perhapsolutely·24 Mar

@elliott__potter What I’ve not been able to set up so far (surprisingly) is a system where an agent can make phone calls with some guidelines and goals, and report back when it’s done. Is that something I can do with your system?

English

Elliott Potter@elliott__potter·23 Mar

We’ve raised $27M for this moment: starting today, your agent gets an iPhone and can talk like a friend. Texting is the universal interface. Billions of people text every day, but until now, developers have been restricted from building on the most powerful channel to ever exist. Linq is a single API for iMessage, RCS, SMS, voice, and even FaceTime and Find My. Nothing for users to download. Nothing new to learn. We’re already powering @interaction, @pika_labs, @getlindy, @zocomputer, @joindimension, Tomo (and others we can’t name just yet) to bring this new ecosystem to life. Join them, and start building for free in our sandbox, linked below. Or comment and we’ll get you set up.

English

389

186

1.8K

723.9K

Entdecken

@thsottiaux @romainhuet @AriX @ZackerySWard @notiansans @runwayml @rohanvarma @rileybrown