Connor Clements | Marketer

@bymarcoperez Note: I built a tool that i use, just a website basically, you can do the same in 5 minutes, super easy to do This will speed it up, you can literally just drag 20 videos in, send it off, have it auto add them to your CSV

@bymarcoperez Im in a different industry, but ive found, getting Claude to generate a detailed prompt, to give Gemini to analyze videos. Generate that description, give it to claude to log in either a MD file or a CSV ( I do CSV ) And ONLY feed it VIRAL videos

Recreated this snapchat hook + app demo format Lets see how it performs

I would say, there's just not enough context I tell Claude to treat each video like a gun-to-your-head, and a drug dealer its important you get people hooked is essentially what im trying to convey Watching the video, i had no idea what the app does, but most likely people didnt get that far Again sounds harsh, not meant to be

@ClementsCopy Thanks for the honest feedback. I would say the algo agrees with you because it didn’t perform well. Going to try some different formats

@bymarcoperez You live on twitter as much as me, that was a fast asf reply lol

@bymarcoperez And then I have it build a learnings.md file, based on the large descriptions of the videos, Gemini gave it in the CSV My next step would be to feed it videos that didnt go viral, so it can mark them as bad

@stephenbtl Does Flux Klein have inherent blocks on making chest size bigger. Ive trained now 12 lora's, trying to do this, and my results are super scattered Training this model is hard, or maybe im just hitting a deep-rooted block

@bridgemindai Steaming today? Or we gonna get a video out of it?

Composer 2 Fast is the move. Same intelligence as Composer 2. 3x faster. $1.50/M input. $7.50/M output. Still cheaper than Claude Opus 4.6 and GPT 5.4. Speed wins in vibe coding.

@jackfriks Wtf its massive

okay i bought it. now i try for 1 week and if i dont love it ill return it

@Emarky Weren’t you in Jack Kitchener’s course? I swear I saw you on Facebook years ago lol

This video was edited entirely by Claude Code. I just gave it the files. If 50 people comment, I'll share the exact setup.

@DavidOndrej1 Didn’t @bridgemindai test this and it wasn’t good at all?

MiniMax just dropped M2.7 and it's approaching Claude Opus 4.6 on real engineering benchmarks. At $0.30/M input and $1.20/M output vs Opus at $5/$25. That's roughly 17x cheaper on input and 21x cheaper on output. Here's what's actually going on 🧵

@OfficialLoganK @aki_ranin Please respond to BridgeMind, Logan

@aki_ranin :)

Hallelujah @OfficialLoganK

@bridgemindai @OfficialLoganK Whatcha think? Matt's been building in public for ages now, streaming most days, and he has constant issues with Gemini on his streams

Just cancelled my $250/month Google AI Ultra subscription. Gemini 3.1 Pro is inconsistent. Gemini CLI crashes mid-session. Antigravity loses context on complex tasks. $250/month for a model I can't rely on. Not worth it. Claude Opus 4.6 and GPT 5.4 stay on the stack. Gemini doesn't.

@bridgemindai @claudeai my mate in AUS said he was having it yesterday but not today

@bridgemindai @claudeai Must admit, ive been using it for the last 8 hours, ive not had it once, weird

Why do I keep having to log in to Claude Code Fix this @claudeai

@OfficialLoganK @vincos @GoogleAIStudio Will we ever see the censorship reduced? Like, it feels pretty nuts we cant turn a picture of something into wearing a bikini. A bikini (even within context at the beach), is crazy work

@vincos @GoogleAIStudio Fix coming soon! We are adding more controls to these deployments which should land this week.

Why am I constantly getting this message in AI Studio lately? @OfficialLoganK @GoogleAIStudio

@ivanka_humeniuk @higgsfield @StefanGeorgi like this - "$100M Netflix budget" - blocked

Who needs a multi-million dollar film crew when you’ve got Seedance 2.0 and @higgsfield SOUL 2.0? 🎬🔥 Tbh, I’m literally sitting here in my oversized hoodie, drinking 3-hour-old coffee, while creating cinematic masterpieces that look like a $100M Netflix budget. The way these models handle textures and movement now is just insane. 🤯 I mean, look at that katana detail! If you told me last year AI would do this, I’d say you’re capping. 🧢

@bridgemindai oooh i thought i saw on stream you tell claude to make a viral tweet template for something

@ClementsCopy No AI on the tweets. I just write what I'm building every day. That's it.

My vibe coding workflow right now: Claude Opus 4.6 builds. GPT 5.4 High Fast reviews. One model ships the code. The other audits everything before it goes live. Two frontier models working together in BridgeSpace. This is how I ship every day. Vibe coding isn't just prompting one model anymore.

@JoshAntBrown @PaulRBerg Yep, hundred percent, i agree it feels better for me too

@PaulRBerg I might be in a minority - but I’ve come to like this. Most of the time I’m dumping stuff in there for context and I don’t really care for finessing it. Claude can figure that out. I can then focus on the stuff above and below the thing I’ve pasted.

This is the most annoying thing in Claude Code. Hiding raw text when you paste more than 4 lines. Terrible UX decision.