Seedance 2.5 is live inside eComrads.
We've been testing it in the engine, ahead of the public API. It's the biggest jump ad video has taken since we started:
• 30-second ads in a single take, no stitching
• sound generated in the same pass as the picture
• 50 references per generation, so your product, your face, your sound stay accurate
• edit one region of one frame, keep the rest
Everything in this film is Seedance 2.5, generated inside eComrads
Seedance 2.5 generates. AdDNA knows what sells.
@PromotingAI Interesting take. The winners probably won’t be the ones with the best tools, but the ones who know how to apply them to create real outcomes
It's been a week of endless nights, alongside keeping up with the fastest week in AI, Seedream 5.0, Nano Banana Lite, and a dozen other content models dropping. In the middle of that, we me and danny ran a 12-hour workshop over several days just to nail our "ideal" ICP (that's how fast things are moving, literally) and our marketing funnel. But somewhere in those 12 hours we figured out a lot more than that. Sharing a few here, might be useful to founders starting out like me:
1- Services: The New Software, published by Sequoia. Same thesis YC wrote into its RFS for AI-Native Service Companies: sell the work, not the tool. Services spend dwarfs software spend.
2- everyone heard about Forward Deployed Engineer so we digged more on the AI native service firm angle and maybe danny just coined a term idk "Forward-deployed marketers" (from a16z's services-led-growth / FDE reading). so his idea was AI does the intelligence, humans hold the judgment that's how you get a real, valuable asset instead of slop. Raw generation is near-free. Taste, prompting craft, knowing what converts AI alone can't do that yet.
3- The demo-vs-delivery gap isn't a bug, it's the business. Every AI content creation tool Higgsfield, Arcads, Creatify, us shows a demo it can't reproduce without a skilled operator behind it. AI can't hand you the end result unless someone knows how to steer it: imagining the scene, communicating it to the model, knowing the model's quirks, running retries without burning the budget and also content generation has gotten expensive fast. It really comes down to how well the operator knows the tools, and that's exactly the team we have. Whoever supplies that operator captures this rising oppuritnity because all the marketing agencies already using AI for content/ads are either trying it themselves or commenting under every AI Infulenecer post "prompt" and then keeping the library of prompts but things are moving rather fast everyone needs an AI partner either to scale their operations or upgarde their current team, especially AI startups are now realizing marketing is very cruicial has to be in-house will pay for a team that hand-holds them for 3 months until they're proficient, then stays on as an advisor.
4- while trying to figure out if our products PMF this came across "Elena Verna (Lovable, Head of Growth): PMF is no longer a static milestone it's an ever-evolving moving target". its crazy that feature differentiation is dead and the only moat that is left beside the endless VIRAL marketing is the trust on humans responsbility you cant BLAME "AI" and also continuous human-centric iteration so we have come to the point of SINGULARITY not Human not AI not Organic not Paid its EVERYTHING and ALL at once - so we are still chasing the ever evolving moving target.
5- 75% of HubSpot's revenue comes from partners (SaaStr). Proof that services distribution scales inside a software company.
anyways 12 hours in, and we only got to solve our top of the funnel that is "the web-app is the demo and MCP is where the "aha" moment actually lives" Monday tomorrow excited and tensed at same time for our next workshop it really is best time in history to be building anything.
Same brief. Same product. Two frontier models.
Seedream 5.0 Pro on top. GPT Image 2 below. Both came back as finished match-day ads inside eComrads.
New models upgrade the engine. Your creative carries over.
Which one would you run?
Seedream 5.0 Pro is now live in eComrads.
• Mark up a photo, get the exact edit
• Rough sketch in, finished ad out
• One ad, six languages, type stays sharp
Every output scored by AdDNA, trained on what converts.
A winning ad account runs on new creatives. The testing never stops, and no team shoots that fast.
We connected eComrads to Claude. On Claude's new Fable 5, one typed sentence returns finished UGC ads.
The five in this video took minutes.
Fable 5 vs Sonnet 5 vs Opus 4.8 on the same product ad.
All three ran an agentic pipeline through eComrads MCP on Claude: read the product, write the creative direction, render the finished ad.
Same brief, but each model directed it completely differently.
Show Claude a viral ad and it builds the same one for your product.
It reads the video frame by frame, breaks down every shot, and renders a finished ad with your product kept accurate.
The format that went viral? recreate for your product with eComrads AdRecreate.
Got tired of watching people manually:
study competitor ads
rewrite hooks
brief creatives
generate UGC
review outputs
We built eComrads CLI with Skills.
Each skill = one job
Together = a full campaign system
Now you just describe your product once
and it handles the rest
from research to ready-to-run ads
Making content was never the hard part. Knowing what sells is.
Introducing eComrads CLI + Skills.
Run eComrads right from your computer. Describe your product in one line, in plain words, and it builds the whole thing. The photoshoots, the videos, the UGC, the static ads. Your product kept accurate instead of reinvented, every output checked before it ships.
We built it as skills. Each skill is one job eComrads knows how to do, the product photoshoot, the videos, the UGC, the static ads, the competitor research, the ad rebuilds. They load together and run as one, so a single instruction becomes a finished campaign.
eComrads studies what wins in your niche, the competitor ads, the hook patterns, the AdDNA, then builds toward what converts, not what looks good.
One command. Every skill. A whole campaign ready to ship.
Introducing eComrads MCP.
Describe your store in one chat, in whatever AI agent you already use, Claude, Codex, any client that speaks MCP, and eComrads builds it. The storefront, the photos, the videos, the UGC, the ads. Meta-ready.
Making content was never the hard part. Knowing what sells is. So before it builds anything, eComrads researches what wins in your niche, the competitor ads, the hook patterns, the AdDNA, then builds toward what converts, not what looks good.
@eComrads As one of the developers behind eComrads MCP, this is exactly what we built it for turning natural language store intent into production-ready storefronts and ad assets grounded in real performance data, not guesswork.
@eComrads Spent a lot of time making ads for brands.
One thing stood out:
The best-performing ads usually aren’t the most polished ones.
They just hit something.
A pattern, an angle, a trigger that makes people act.
That’s what eComrads is built around.
We're excited to introduce eComrads.
Our mission is to make AI that sells, not just AI that makes content. We're building AdDNA, the intelligence layer that gives AI models and agents judgment about what actually converts.
And today, eComrads is live to everyone.
AI has made it easy to generate anything. A product photo, a video, an ad, in seconds. Everyone has the same tools now, and most of what comes out is slop, content that looks fine and sells nothing. Because generating content was never the hard part. The hard part is judgement. What converts, what fits the brand, what makes someone actually buy. Cracking that means turning something fuzzy and subjective, the instinct for what sells, into something we can measure and codify. We're starting with ecommerce.
There are two sides to this, the intelligence layer and the production layer:
- We've trained AdDNA on what actually performs in each niche, the winning patterns behind real conversions, not pretty pictures, and we refresh it constantly so the system generates toward what sells.
- We've built the production engine that turns that judgment into finished, on-brand creative, photoshoots, video, UGC, and ads, with your product kept accurate instead of reinvented, and every output checked before it ships.
We learned all of this the hard way, producing creative by hand for real brands until we knew exactly what breaks and what sells. eComrads is that knowledge, turned into a product.
We want a future where any brand can compete on the quality of its ideas, not the size of its ad budget.