Samuel Roy | Contextual AI, macOS, Founder

1.7K posts

Samuel Roy | Contextual AI, macOS, Founder

@samuelroy_

Building Alter. Contextual AI for macOS. Sharing lessons on AI, product, and scaling. 1000+ member community. Join me: https://t.co/4AhID7OvMx

France Присоединился Aralık 2012

143 Подписки152 Подписчики

Закреплённый твит

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·5 May

Codex pets are escaping! We built OpenPets so every local agent can meet in the same little desktop world: Codex, Claude Code, Cursor, OpenCode, Pi, MCP clients, apps, and scripts. One shared pet. Many tools. Running, reviewing, waiting, celebrating. Open source, Swift 6, MIT. github.com/alterhq/openpe…

OpenAI Developers@OpenAIDevs

Pets. Now in Codex. Use /pet to wake your pet.

English

397

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·1d

Cool to see OpenAI exploring this direction. AppSense in Alter has been doing this for the past year: app-aware AI context from any window, with support for both frontier and local models. Link: alterhq.com youtu.be/xbERBld3aP4?t=…

YouTube

OpenAI Developers@OpenAIDevs

It’s Codex Thursday, and yes, we have updates for you. First up: Appshots, a new way to bring the context of what you’re working on into Codex. On your Mac, press Command-Command to attach your app window to a Codex thread. Codex gets both a screenshot and text from the window, including content beyond what’s visible onscreen. Appshots are available across plans on Mac, with enterprise access coming soon.

English

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·3d

Now that we have the Claude prompt cache diagnostics, who's willing to share some data? I start: 1. Overall Cache read ratio at 98.9% 2. 80% of cache miss are due to messages changed. It's super surprising, we designed the Alter harness to avoid that. The team is looking into it, I will share more insights if it's interesting. 3. Write amort. for Sonnet is at 3.69x

Samuel Roy | Contextual AI, macOS, Founder tweet media

ClaudeDevs@ClaudeDevs

Prompt cache diagnostics are now in Claude Console. When a request misses the cache, you can now see exactly which part of your prompt changed and how many tokens it cost you.

English

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·3d

@kudnever @NousResearch @grok @grok what are the usage limits of Supergrok

English

kudnever@kudnever·3d

@samuelroy_ @NousResearch @grok do u know smth about limits for grok with high reas? are they alr?

English

kudnever@kudnever·3d

gm finally setted up my own AI agent using Hermes @NousResearch > @grok module for X search (freshest info on everything) > polymarket MCP server > Codex API as LLM for Hermes (5.3-codex) > GitHub OAuth to push, merge my code am i missing smth for my setup?

English

227

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·3d

@max_bevza They are all here

English

Max Bevza 🦆 Wallchain Quacks@max_bevza·3d

ChatGPT gets 8x more Google searches than Claude Claude users, where are you hiding?

Max Bevza 🦆 Wallchain Quacks tweet media

English

1.7K

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·3d

@kudnever @NousResearch @grok I think it's worth trying with reasoning effort at high

English

kudnever@kudnever·3d

@samuelroy_ @NousResearch @grok i believe 5.3 codex to be a better model for tool calling and overall coding do u think its worth trying grok?

English

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·3d

@gauravsbuilding Claude Code for X is the new Airbnb

English

Gaurav@gauravsbuilding·3d

Today we're introducing Claude Code for Marketing. In one prompt, Fastlane deploys social media accounts, generates viral content, and posts everything for you automatically. This is beyond insanity.

English

321

346

5.1K

2.5M

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·3d

@senamakel How much do you spend running this much?

English

322

Steven Enamakel@senamakel·3d

I run 18 claude/cursor/gemini instances at the same time and work on 18 different ideas/features/bugs at the same time. And then I have over 15-20 agents running on the cloud. I ship & code like a fucking madman.

English

3.8K

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·3d

@deadlock123123 @nexxeln Tell us when you switch again so we know what to do

English

DEADLOCK@deadlock123123·3d

@nexxeln Claude Code got dumb so I switched to Codex, and the second I got here this one started acting up too..

English

nexxel@nexxeln·4d

why is gpt-5.5 being so dumb rn

English

446

48.2K

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·3d

@Ali_TongyiLab Any chances to get that released through other providers?

English

624

Tongyi Lab@Ali_TongyiLab·3d

1/6 Introducing Qwen3.5-LiveTranslate: Next-gen real-time interpretation is here. 🌍 We’re breaking down language barriers with 3,500+ language pairs, ultra-low latency, visual context, real-time voice cloning, and hotword customization. Engineered to help you ship native, frictionless real-time translation experiences to a global audience.

English

697

4.1M

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·4d

@Andy_AJT If you open-source your project, I'd love to add it to github.com/alterhq/awesom…

English

Andy T@Andy_AJT·4d

Made codex-pet-bar He even waves at you with a little codex flag when you need to approve something.

Andy T@Andy_AJT

I put my Codex Pet in my macbook MenuBar. And of course hes a Goblin. @OpenAIDevs

English

2.9K

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·4d

More than 3500 Codex Pets ready to use What's your favorite @OpenAIDevs? openpets.sh/?sort=new

English

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·4d

@joshua_soileau @ParkerRex That's smart!

English

Joshua Soileau@joshua_soileau·4d

@samuelroy_ @ParkerRex If you use the Codex desktop app, ask it to put the images into an HTML page, open it in the in-app browser, and you can use the annotation tool in Codex to point out specific spots of the image and comment on them, and Codex will understand what you're pointing at

English

Parker Rex@ParkerRex·4d

Ask Codex to generate balsamiq style mockups ahead of any long-horizon tasks. You end up saving so much time going through the wireframes, mapping out all the states, all the views, all the affordances and signifiers. Then take those and do the visual pass. Then when you're running Exec Plans or Programs, it has this to reference.

English

386

20.3K

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·4d

@mosesbabychrist Is it the beginning of a new social platform? 😅

English

MBC@mosesbabychrist·4d

Our community-powered gallery of /pet's is growing stronger every day! And now you can comment on uploads, and see recent discussions. codex-pets.net

English

143

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·4d

@ParkerRex I imagined annotations directly on the mockups similar to what you can achieve with the in browser for better accuracy

English

Parker Rex@ParkerRex·4d

@samuelroy_ You iterate the mockup in Codex just send a bulleted list of what to change and do another round of wires

English

401

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·4d

@jumperz Discord is the underrated AI orchestrator

English

246

JUMPERZ@jumperz·4d

this setup is too clean honestly.. grok pulls x signal raw, hermes verifies it into notes with sources, discord threads keep every research topic isolated and ofc we have Codex sitting underneath handling whatever needs a cron or automation hook. >codex 5.5, main llm >grok, raw research >hermes , verification + orchestration >discord threads , memory + workspace this is what you really need.. nothing crazy

English

5.6K

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·4d

@fcoury A lovely place for my new favorite Snorty! openpets.sh/pets/snorty

English

768

Felipe Coury 🦀@fcoury·4d

What day is today? Pets on terminal day! 🐾🐕🐇

English

13.9K

Samuel Roy | Contextual AI, macOS, Founder@samuelroy_·4d

@cursor_ai Just when I was thinking about dropping my $200 sub for Codex.

English

Cursor@cursor_ai·4d

Introducing Composer 2.5, our most powerful model yet. It's more intelligent, better at sustained work on long-running tasks, and more reliable at following complex instructions. For the next week, we’re doubling the included usage of the model.