Martin X

@sama Every time I check some of my old codebases I find it hard to believe how I've been able (and had the patience) to write so much code by hand. It's a strange feeling indeed.

I have so much gratitude to people who wrote extremely complex software character-by-character. It already feels difficult to remember how much effort it really took. Thank you for getting us to this point.

The ones I hate most are those that sound like marketing copy and have this typical Chatgpt cadence.. super annoying.. but maybe not trivial to filter reliably because many people use AI to express their thoughts more clearly while genuinely wanting to add value. What they can do is boost manually typed tweets that are reviewed before being posted (a bit of scrolling, some edits here and there like what I'm doing now lol).. this would be very inefficient to automate on real devices and a reliable indicator for human typing. But of course many people use voice input as well that might be auto polished by AI... so definitely not a trivial fix

When are they going to ban AI replies here on X? Totally worthless for everyone and hurts the user experience.

@Legendaryy I agree it's a "deal" for power users, but quite a few barely use their sub. I know tons of people on Chatgpt plus that don't know what is Codex. So I guess it evens out, but indeed it might get more expensive in the future.

Your Claude subscription is massively subsidized and it won't last forever. A $20/mo Pro plan burns through ~$180/mo in API-equivalent tokens. Heavy Max users hit $5K/mo on a $200 plan. Actual compute cost is roughly 10% of API pricing. Venture capital covers the rest. Anthropic just 2x'd Claude Code limits for "Spring Break'" during off hours. Enjoy it while it's here. At some point the math has to math. Clip from the @modernmarket_

@C0NWIC @Legendaryy Makes sense, although the ToS of most AI providers usually claim that they only use free users input for training (or at least non-business users). But what they say and what they do are different things , hehe.

"AI is subsidized" is the best psyop in tech right now. You're paying $200/mo thinking you're getting a deal. You are, but so are they. Every session is live labeled training data: correcting outputs, steering agents, building real systems that no web scrape can touch. Also, the tiers aren't just pricing. They're a data quality filter. $20 users ask trivia. $200 users build products. Guess whose sessions train the next model. They'd make it free if they could. The compute won't let them. So they sell you "subsidized" and collect gold from both ends.

Finally fixed. Thanks @OpenAIDevs

@naval Damn. So vibecasting is dead? It's hard to keep up these days ..

Coding an app is the new starting a podcast.

@OfficialLoganK @googlemaps Is it available globally? I don't see Ask Gemini anywhere in Maps on Android (using latest version)

Introducing our biggest upgrade to @googlemaps since the original launch, featuring Ask Gemini (with personalization), Immersive Navigation, and much more!! 🗺️

@NemTheManX @antigravity I am using Claude Pro and Codex (on Plus sub) as well, their limits now seem to be more generous compared to Google's Gemini which is crazy.

@martogram @antigravity Just switch to Claude. I like the idea of Antigravity - but if Google do not fulfill customer satisfation it feels like not a good investment.

We’re evolving Google AI plans to give you more control over how you build. Every subscription includes built-in AI credits, which can now be used for Antigravity, giving you a seamless path to scale. Google AI Pro is the home for the practical builder, hobbyists, students, and developers who live in the IDE and don't necessarily rely on an agent. This plan features generous limits for Gemini Flash, with a baseline quota included to "taste test" our most advanced premium models. Google AI Ultra serves as the daily driver for those shipping at the highest scale who need consistent, high-volume access to our most complex models. If you’re on Pro but need "extra juice" for a heavy sprint or deeper access to premium models, simply top up your AI credits to customize your plan. Keep building. Keep shipping.

@googledevs @grok is it same price as gemini-embedding-001?

Announcing Gemini Embedding 2 ✨ the first fully multimodal embedding model built on the Gemini architecture. Now available in preview via the Gemini API and Vertex AI. The new model provides semantic understanding across 100+ languages — and support for modalities across text, images, video, audio and documents (PDFs) in a shared vector space. Start building today. Read the blog to learn more: goo.gle/4los3aI

@aryanlabde proper logging and analytics also become very crucial very quickly and are not trivial to implement when you haven't thought about them from the very beginning

Vibe coding a product is so easy. Until you’re hit with: - auth - security - workflow logic - context windows - confirmation emails - payment integration That’s where you wish you knew how to code.

Gmail: “Here’s a full AI-generated email for you.” Also Gmail: “You’ll have to write the subject yourself.” Maybe I’m missing something, but why is that part still manual?

Hey @OpenAIDevs , please fix the Windows VS Code Codex extension file/line link bug. Clicking file references opens a broken file+.vscode-resource.vscode-cdn.net URL in the browser instead of focusing the correct file and line in the editor. You keep shipping new models, but long-standing bugs like this still aren’t fixed. #VSCode #Codex

@ZackKorman Yep.. they are shipping features for banned users while Antigravity is failing on every second prompt with agent errors, no auto-retry whatsoever, I come back after 20 minutes only to see it crashed and did nothing. It's just ridiculous.

“Improved UI for banned users” will be the title of the book on how Google lost the AI era.

.@antigravity A second update in the same day… just to roll out an “improvement” for banned users? Everyone gets restart prompts for something that affects a tiny subset. That kind of rollout is honestly frustrating. #Antigravity

Why does Samsung keep selecting presenters for the opening speech who struggle to speak English? #GalaxyUnpacked

@elbeyoglu Are you using an open source lib to parse the DOM into md?

I built markdown.new Put markdown.new before any URL → get clean Markdown back. Cloudflare's Markdown for Agents is great, but only works for enabled sites. markdown.new works for ANY website on the internet. 80% fewer tokens. Also converts PDFs, images, audio. Free. No signup. markdown.new

I'm starting to like @AnthropicAI more and more youtu.be/De-_wQpKw0s

@chatgpt21 1M context window by itself means nothing. What is in the context matters. And coding agents can get way better at optimizing the input prompt on every turn by removing irrelevant stuff from previous turns and summarizing relevant details proactively, not when the window is full.

OpenAI is slating a 5.3 release sometime in February But the I’m getting word internally that it will not be called sonnet 4.7 and in fact will be called Sonnet 5. It have a 1M token context window & outperform Opus 4.5 on some agentic coding tasks! Remember we are also getting Grok 4.2 & Deepseek V4 February will be a really exciting month! Just like last year!