Noel

@ReflctWillie What sites/app is this?

Call it anecdotal, because it is, but it's working well for me. PROMPT: "Analyze the reference images and catalog all of the minor details of the setting, characters, car placement, signage, the character in the orange suit, and ground textures. Show me a view from a high angle looking down at the car at an angle. The character in the button down shirt is sitting in the car with one arm resting on the top of driver's side door. The man in the orange suit is standing in front of the car in the exact same place as the reference image. Match the color of the reference image perfectly. Match the position of the car in the reference image perfectly."

Tip for Nano Banana Pro if you see it dropping details. "Analyze and catalog all minor details in the reference image, take note of [list the details]..." This activates the reasoning core in NBP to take note of key elements of the image. Context below, thank me later.

@businessbarista Claude

I want to start a community dedicated to Claude Code. It’s become the gateway drug to coding and experiencing the power of AI for tons of people. This will be a space for people to share killer use cases, agentic workflows, proven prompts, and connect with other CC obsessives. Comment “Claude” if you want to join.

@RuffBuffet @wallpapermag @Apple You know you can … just turn it off right? If you’re actually worried about battery life.

@wallpapermag @Apple What’s this mean for Liquid Glass? Can we dump this eyesore and get some performance and battery life back? My Series 10 Watch lost 30% runtime with the jump to OS 26 from 11. Barely survives a day now. Even my 16 Pro Max went from a battery life king to always dead.

.@Apple enters a new chapter in Human Interface Design as Steve Lemay becomes the new leader of the team, working alongside Molly Anderson, head of Industrial Design at Apple. A veteran at the company since 1999, Lemay has shaped every major interface evolution from the original #iPhone to macOS, iOS, iPadOS, watchOS and, most recently, visionOS. Tim Cook calls him “an extraordinarily high bar for excellence,” noting his depth of understanding in platform technologies and his gift for shaping intuitive, human-centred digital experiences. With Lemay now at the helm, Apple’s design bench continues to push forward, refining the interactions, systems and visual languages that define the world’s most recognisable devices.

@_LaurentB @LinusEkenstam Way to bet on the least possible thing to happen. Also it would be awful given how censoring the company is value system wise.

@LinusEkenstam Apple releasing an AI image/video model.

This week will mark a major shift in how we perceive AI video. Things are about to get crazy good. Seemingly over night. It will also come from a place most people are not expecting. Excited for this week.

@vladfaust @fofrAI @Must_be_Ash And or even a note from a single model provider that the models do anything unique with json or yaml or whatever structure people seem to be promoting use of…

@vladfaust @fofrAI @Must_be_Ash I don’t ever see any concrete info that json leads to higher quality. I’ve asked every time I see it brought up. If anything it consumes context - which is bad. If people like it as a “workflow” aid I guess. But “better” quality. I’m still waiting for evidence…

I’ve started a blog. Here’s a guide on prompting Nano Banana Pro: fofr.ai/nano-banana-pr… The blog was made with AI Studio, and Gemini 3 Pro in Cursor.

@Gurhi_jalebi Also bags like this are stupid and unergonomic.

@ArmanHezarkhani Deleted my reply when I realized you’re not talking about image/video models.

XML is adderall for AI. It gets your agents to focus on what matters. Here’s how it works. Why it matters: Most prompts are mush. XML gives your model a contract: clear sections, strict outputs, easy parsing. That means higher compliance, fewer hallucinations, and responses you can wire into code/ops without manual cleanup. What it is: You’re not building an XML app—you’re using tags as “guardrails.” Tags label intent (), inputs (), constraints (), and—most importantly—the exact output shape (). One root element, fixed tag names, no extra prose. Models follow it astonishingly well. How to use it: Reply to this post with "XML Guide" for my guide on how to use XML in your prompts.

@carlvellotti CC

Every PM should be using Claude Code. So I built a HUGE course for you to learn Claude Code... IN Claude Code! 🔹 Complete guide 🔹 Make PRDs, analyze data, create decks Soon, I'll sell it for $149. For the next 24h: FREE! Follow + RT + comment "CC" & I'll DM it.

@ChaceMcLees @chatgpt21 @emanuelturis @Casey The AI slop being addictive is completely different point than the sidewalk metaphor and stopping though… those are diff topics altogether.

@chatgpt21 @emanuelturis @Casey Except accidents aren’t engineered specifically to maximize engagement with my senses. Nobody trained people on how to wreck cars based on my preferences for viewing accidents. This is social engineering powered and driven by AI slop. It’s designed to make us want to look.

my entire Instagram feed is Sora clips

@rabois @brackin @nejatian Opendoor is barely even SF. What are you two talking about?

@brackin @nejatian Even better.

Opendoor returns to FounderMode. And we just hired the absolute best executive who has a founder brain as CEO: @nejatian.

@Brick_Suit @J_Swish24 I agree w Brick that video itself was obnoxious and dumb.

@J_Swish24 Bro, you can't even do a freeze frame. Get rekt.

Forget the D list influencers there strictly for the instagram pic. These are the worst kind of people at the US Open. Johnny ear piece. 🤦🏻

@thekitze hyperlitemountaingear.com/products/daybr…

I wanna become the person who flies with only a backpack (not even carry on) anyone doing this? which backpack?

@mrjeremyblaze Now try to do anything involving a Boolean in a way Jira could do a decade ago…

The sheer beauty and simplicity of Linear's AI filters blows my mind. It's such a tiny detail but packs a huge punch.

@amoufarek @philipjohnball Go check them out where? In this post? Or is there a site for these now?

Our team has been experimenting with all sorts of real-time and interactive generations with Genie 3, please go check them out! A great one, hot out of the oven from @philipjohnball

@samcamachoh @imPatrickT @b_hvrd Why not … you know buy a wireless microphone?

@imPatrickT @b_hvrd Smart

Just got mine. What should I know

@philchacko @cnnrjcbsn @pitdesi @meetgranola The Granola transcribed does not run locally. granola.ai/docs/docs/101/…

@cnnrjcbsn @pitdesi @meetgranola Personally I don't consider having a text transcript any creepier than typing into a Google doc during a meeting, since the transcriber runs locally. Very different than producing and saving a raw mp3 or mp4.

scary thing i've heard happening far too often with notetakers very happy to be using laptop-side @meetgranola

@jerkeyray Imagine sharing a link which is useful vs a screen shot which is work.

dude just built something which made all web scrapers obsolete

@midjourney @kriebe Top reply is a simple - check your math - sorry. 😆

@kriebe oh sorry yah it's like roughly 4x it's 720p

Today we're releasing our *HD* Video mode to Pro and Mega subscribers. It's ~3.2x more expensive with ~4x more pixels than our default SD video outputs. This is for professionals that need the absolute highest quality footage possible out of Midjourney and we hope you enjoy it.

@ireallyhateyou @SkipTheT Outputting certain amounts of specific characters at least pre reasoning model was very very difficult for LLM to do.

@SkipTheT I recently asked ChatGPT to shorten a text to make it shorter than Twitter's character limit. It repeatedly kept issuing loner-than-tweet texts and claiming they contain less than the amount of characters they actually contained. After 4 attempts I just did it myself.

After insisting an emaciated old man in Gaza is "AI fake" Grok finally admits its mistake. Never trust AI chatbots blindly.