MindMirror

@pvncher @KingBootoshi @TheRealAdamG @RepoPrompt This is output from a 150k token [review] prompt from RepoPrompt that I just got seconds ago in 5.5 pro

@TheMysteryDrop @KingBootoshi @TheRealAdamG Works well for me with prompts made in @RepoPrompt

gpt 5.5 pro is an absolute monster btw not codex xhigh, im talking about Pro in the web interface you should be talking security/backend/infra/design with this one, giving your full files and max context Pro has been an absolute Godsend ever since we finally got 1m context in it. It was only a couple months ago we can only fit like 60k tokens in context, now I’m pushing ~250k with EASE (btw don’t go past ~250k, 1m context does not mean stuff it with literal 1m context because it gets real bad accuracy past this point) it doesn’t seem to over engineer either, which was a big issue with the other models it’s soooo good for backend work btw, do not continue a prompt chain with pro. each message should essentially be a one and done. if you want to follow up reconstruct the context from scratch!

@pvncher @KingBootoshi @TheRealAdamG I find that it’s way less than 270k. Even at 150k with instructions I usually get something like “you just sent me a bunch of code. What do you want me to do?”

@KingBootoshi @TheRealAdamG You don’t get 1m context on the web fyi. It’s 270k. More than that and the model has to selectively read your files

@mikefutia Statics

I just vibe coded a static ad generator in Claude Code that creates 100+ Meta ads in minutes. All using the new, insane ChatGPT Images 2.0 model. One competitor ad + your product photo + your brand kit = dozens of on-brand variations, each targeting a different customer persona. Built 100% in Claude Code on the new ChatGPT Images 2.0. Perfect for DTC brands and agencies who need more statics at scale. Here's how it works: → Upload any competitor ad as your reference template → Add your product photos and brand kit (colors, fonts, logos) → AI generates 10 customer profiles from your brand research → Pick how many variations you want (10, 20, 30) → Tool fires every prompt to ChatGPT Images 2.0 with persona-specific copy for each one No designer back-and-forth. No Canva templates. No generic "Shop Now" on everything. What you get: → Ads that mirror winning concepts in your brand's voice → Text that actually renders correctly (the new model handles dense copy, logos, and multi-language callouts cleanly) → Copy targeted to specific customer pain points and personas → Multi-brand/client support with saved brand kits → Reusable customer profiles you build once and generate from forever I recorded a full walkthrough showing exactly how this works, including ALL the prompts I used so you can build it yourself. Want access to all the prompts for free? > Like this post > Comment "STATICS" And I'll send it over (must be following so I can DM)

@ronaldmannak Just use a proxy to pass your Claude subscription. This one works well with Droid: github.com/automazeio/vib…

Now for the bigger point. What I really want to see next is affordability. Yesterday, Factory’s CEO hinted that Missions will become more affordable. I don’t know whether that means smarter model routing, better token efficiency, or something else entirely, but it would be a very welcome improvement. Because right now, I simply can’t justify spending $200 a night. x.com/matanSF/status…

Yesterday’s tweet about Droid Missions clearly resonated, so here’s the fuller story how I used Missions and how it completely changed how I think about vibe coding. x.com/ronaldmannak/s…

@levelsio They're not dead. I am testing them across a few agency accounts. FWIW, their AI creative is very low-quality, but ironically they offer good value for human UGC.

Icon, the AI Admaker, just went bankrupt They paid $12M for the domain Icon.com and now it's dead

@AdamHoltererer @oztsori The challenge was “Try to prompt this” - not “Try to create a work as original as mine” Nothing can generate a perfect replica of an image based on text alone because text is far too lossy. It’s not really a coherent challenge or measure of AI.

@TheMysteryDrop @oztsori The point of the post was that this type of original and highly personal art isn't something that can be produced from scratch by AI. I think any use of the original artwork as a reference is kind of cheating, but that's just my interpretation.

Go ahead. Try to prompt this. I’ll wait. ☕

@AdamHoltererer @oztsori Not denying it - I used a combination of the prompt and reference images

@TheMysteryDrop @oztsori I just ran that prompt and it still comes out with generic icons and doesn't have the same background pattern. You used his original images as a reference when generating those. Do you deny this? Because if you're saying that you only provided a text prompt, then you're lying.

A 4x5 grid of twenty distinct, textured, hand-printed block-print icons. The background is an alternating warm-cream and charcoal-black checkerboard grid of perfectly aligned squares. The art style is artisanal and tactile, with pervasive ink texture, grit, and soft, imperfect edges for all figures and blocks. Icons are arranged in 4 columns and 5 rows. On cream background blocks, there is a central, charcoal-black textured sub-square. On black background blocks, the element is placed directly on the surface. All dots are small and located in the upper corners of the dark field. Row 1: (Left to Right) (On Cream) Dark sub-square with a white-textured serpent coiled around its S-shape and a green dot. (On Black) White-textured dove carrying a branch and an orange dot. (On Cream) Dark sub-square with a white-textured puffin with a fish and a purple dot. (On Black) White-textured rhino and an orange dot. Row 2: (On Black) White-textured standing dromedary camel and an orange dot. (On Cream) Dark sub-square with a white-textured standing antelope and a purple dot. (On Black) White-textured lion and a red-orange dot. (On Cream) Unique dark-textured square with black-textured zebra stripe pattern on a white field and a green dot. Row 3: (On Cream) Dark sub-square with a white-textured fox silhouette and a green dot. (On Black) White-textured horse's head profile and an orange dot. (On Cream) Dark sub-square with a white-textured small bird perched on a branch and a purple dot. (On Black) White-textured llama silhouette and an orange dot. Row 4: (On Black) White-textured standing giraffe and an orange dot at the top center. (On Cream) Dark sub-square with a white-textured elephant and a green dot. (On Black) White-textured small bird on a branch and a red-orange dot. (On Cream) Dark sub-square with a white-textured long-tailed cat-like creature and a purple dot. Row 5: (On Cream) Dark sub-square with a white-textured polar bear silhouette and a purple dot. (On Black) White-textured rooster and a red-orange dot. (On Cream) Dark sub-square with a white-textured bighorn ram and a purple dot. (On Black) White-textured rabbit/hare and an orange dot. All animal figures are simplified silhouetted forms with heavy texture. Small colored dots are solid. Lighting is even. The tactile quality of a linocut or block print is pervasive.

@TheMysteryDrop @oztsori That doesn't really count because you provided his existing image as a reference.

A 4x5 grid of twenty distinct, textured, hand-printed block-print icons. The background is an alternating warm-cream and charcoal-black checkerboard grid of perfectly aligned squares. The art style is artisanal and tactile, with pervasive ink texture, grit, and soft, imperfect edges for all figures and blocks. Icons are arranged in 4 columns and 5 rows. On cream background blocks, there is a central, charcoal-black textured sub-square. On black background blocks, the element is placed directly on the surface. All dots are small and located in the upper corners of the dark field. Row 1: (Left to Right) (On Cream) Dark sub-square with a white-textured serpent coiled around its S-shape and a green dot. (On Black) White-textured dove carrying a branch and an orange dot. (On Cream) Dark sub-square with a white-textured puffin with a fish and a purple dot. (On Black) White-textured rhino and an orange dot. Row 2: (On Black) White-textured standing dromedary camel and an orange dot. (On Cream) Dark sub-square with a white-textured standing antelope and a purple dot. (On Black) White-textured lion and a red-orange dot. (On Cream) Unique dark-textured square with black-textured zebra stripe pattern on a white field and a green dot. Row 3: (On Cream) Dark sub-square with a white-textured fox silhouette and a green dot. (On Black) White-textured horse's head profile and an orange dot. (On Cream) Dark sub-square with a white-textured small bird perched on a branch and a purple dot. (On Black) White-textured llama silhouette and an orange dot. Row 4: (On Black) White-textured standing giraffe and an orange dot at the top center. (On Cream) Dark sub-square with a white-textured elephant and a green dot. (On Black) White-textured small bird on a branch and a red-orange dot. (On Cream) Dark sub-square with a white-textured long-tailed cat-like creature and a purple dot. Row 5: (On Cream) Dark sub-square with a white-textured polar bear silhouette and a purple dot. (On Black) White-textured rooster and a red-orange dot. (On Cream) Dark sub-square with a white-textured bighorn ram and a purple dot. (On Black) White-textured rabbit/hare and an orange dot. All animal figures are simplified silhouetted forms with heavy texture. Small colored dots are solid. Lighting is even. The tactile quality of a linocut or block print is pervasive.

@TheMysteryDrop @oztsori Write the prompt

@codyschneider Generator

I just had Claude Code build me a Facebook ad generator that can make 100+ on-brand ad variations in minutes for $0. And I made a full Notion document guide for you. It includes: 1. How to use Claude to find the pain points and desired outcomes of your ICP 2. How to use these pain points and outcomes to write ad copy variations 3. How to build a Facebook ad template entirely with code (just like the ones you see) 4. How focus Claude Code’s design so the ad feels “on-brand” 5. How to export the Facebook ads as PNGs in a zip file 6. How to bulk upload them to a Facebook ad set 7. How to use an AI data analyst to track the success of these ads Everything above is just API calls and Claude Code doing the work for you. You just come up with the ideas and polish the outputs. Like and comment "generator" and I'll send the Notion document to you

@tankots @WisprFlow Whispr Flow

We offered 5 people a Porsche 911 GT3 RS if they could get @WisprFlow to make a mistake It's the fastest and most accurate AI voice dictation app that's 3x more accurate than ChatGPT, Claude, or Siri. Today, we’re finally launching on Android. Download now: play.google.com/store/apps/det… As a part of the launch, we’re giving away 6 months of Wispr Flow Pro for free. Like, retweet and comment ‘Wispr Flow’ to get it. Enjoy. — Written with Wispr Flow

@nateliason @FelixCraftAI HeyGen API is not a bad option

I have done zero AI video but want to give @FelixCraftAI more ways to communicate with us. What models should we be using? Quality is more important than cost.

@mikefutia @thedanielokon Meta

Meta just dropped the biggest Ad Library update ever: you can now sort any brand's ads by highest impressions 🤯 Which means you can see exactly which competitor ads are getting the most views. @thedanielokon from ACTIV said "if someone can create an AI scraper to review the top 10 ads every week, I'd pay for that." So I built it with Claude Code. Here's how it works: → Import any brand with their Ad Library URL → Apify scrapes their top-performing ads automatically → Click into any ad for full breakdown (headline, copy, CTA) → Gemini watches the video or analyzes the image → Returns asset type, visual format, messaging angle, hook tactic, offer type No manual research, no watching videos one by one, no messy spreadsheets. Here's what I built: - 50+ DTC brands already loaded (AG1, Caraway, Chomps, Dr. Squatch, Gruns, Jones Road, Magic Spoon, Ridge) - Weekly auto-scrape to refresh top ads - AI analysis on any ad in one click - Bookmark system to save winners - Filter by brand, category, media type, or AI tags Built 100% in Claude Code, hosted on Replit. I recorded a full step-by-step showing exactly how I built this. Want to build your own version? I'm giving away the prompts I used + the Claude Code starter template. To get access for free: > Like this post > Comment "META" And I'll send it over (must be following so I can DM)

@yacineMTB @doodlestein The command can be to invoke headless codex for /review

@TheMysteryDrop @doodlestein actually, instead i can just use codex because its miles better

after using codex again, the amount of bugs its finding left behind claude are astounding. it's literally because it lied to me. it actually just didn't follow instructions in a seaky way trying to trick me. what the fuck like actually. i am a hawk. it still snuck shit in

@doodlestein @yacineMTB You can create a command / skill / subagent for this and just bake it into the global Claude.md and it will run this at the end of every task

You can get Claude to reliably clean up its own messes if you just repeat this prompt 3 to 5 times after it writes some code for you (I do this hundreds of times a day): Great, now I want you to carefully read over all of the new code you just wrote and other existing code you just modified with "fresh eyes" looking super carefully for any obvious bugs, errors, problems, issues, confusion, etc. Carefully fix anything you uncover.

@irl_danB Isn't Claude Code also a bet that vertical integration will give them better training data to stay on the bleeding edge re: the best models?

Anthropic’s API bet is that they will always have the best models Anthropic’s Claude Code / App bet is that models will commoditize, and they will need to own the UI or win through vertical integration they cannot serve both masters forever to avoid kerfuffles like this one

@deredleritt3r I’m not saying that OpenAI can’t do it - just that their consumer business seems like an albatross re: their ability to achieve this goal.

OpenAI is planning to devote "hundreds of thousands" of GPUs specifically to the AI research intern this year. There should be enough compute, assuming that the data center build-out goes well. I'm not quite as clear on where the competitors are and the amount of resources they would be willing to devote to such endeavors.

Mark Chen summarizes the goal of OpenAI's research roadmap: "We’re very excited about our 2026 roadmap and advancing work toward an automated scientist." Jakub Pachocki said the same thing in September (see below). Consumer applications are not the goal. Slop videos are not the goal. The goal is to fully automate AI research. For those counting, September 2026 is now just 8 months away.

@kodjima33 code.claude.com/docs/en/checkp…

@kodjima33 Use git + --dangerously-skip-permissions Can also use /rewind or esc to revert