Alex Reiswich

Everyone’s talking about build → deploy → QA loops But the real leverage is BEFORE that: • define the problem clearly • design the system upfront • set constraints + success criteria Because AI can build fast… but it will just build the wrong thing faster if you don’t think first. Save this.

@higgsfield Real question: If everyone can generate content 24/7… what actually makes content valuable anymore?

Higgsfield MCP is HERE! 🧩 You can now create content end-to-end inside any agent: OpenClaw, Hermes Agent, NemoClaw. The only way to get agentic access to Seedance 2.0, GPT Images 2.0, and every other top model. Let your agents build content while you sleep.

5/ What to do now: stop optimizing your output speed. Optimize your ability to spot where agents are about to hallucinate, break a rule, or miss context only a human would have. That skill is the 10%. Everything else is commodity by 2030.

4/ The trap: your performance review still says 'produced X deliverables.' Your actual contribution is 'caught 3 things an agent missed that would have cost us $2M.' Companies that don't rewrite the role around judgment will overpay for production and underpay for supervision.

1/ By 2030, most knowledge-work roles will still have a human in the loop. But that human will be doing 10% of the work they do today. The rest gets absorbed by agents. The org chart survives. The job description doesn't.

@bridgemindai I really would like to know what kind of prompts you are using. I have built a whole working SaaS within an 6 hour session and have like 25% usage. Opus 4.7 feels a lot faster and really intelligent, even UI is by worlds better than 4.6

Claude Opus 4.7 eats tokens for breakfast. 3 prompts. 13% usage. This is why I bought two Max plans. One $200/month subscription would be gone before lunch. The model is the best I've ever used. But you can feel that 35% token increase on every single prompt. Claude Opus 4.7 is not a model you run on a budget.

@ipivto @IXITimmyIXI I think what Ilya was asking was a manual how to get this crazy result 😃

@IXITimmyIXI @AlexReiswich Do you mean you prompted all camera movements (through voice input)?

I tried Seedance 2.0 on Runway and it's amazin' 🤯 Audio-driven AI music videos will be the next wave

@IXITimmyIXI No way. If true we will be flooded by slop in 3 , 2 , 1 lol

@IXITimmyIXI Is this plain lipsync with a suno song and some cutting? I still remember the first video of will smith eating pasta… just amazing

@higgsfield @higgsfield I am now trying already 30 minutes to get a Video finished. It either gets marked as NSFW (my prompts do not have that) or I get other errors. Do you mind sharing any tips for prompts?

Introducing Higgsfield Marketing Studio powered by Seedance 2.0: a single workflow from product to finished ad. Bring a product image or paste a product link, use an avatar from your photo or build one inside the tool, and generate your ad end-to-end in 9 ad styles - UGC, unboxing, reviews, and more. New users get up to 70% off, with up to 14 days of Unlimited generations.* Start here: higgsfield.ai

stop building in stealth. nobody's gonna steal your half-baked idea. they're too busy stealing someone else's half-baked idea.

@KaiXCreator Did you try Hermes by @NousResearch? From my view it’s working a lot better with less maintenance that OpenClaw would require.

The more I use OpenClaw, the more I think it's not just hype. The only thing limiting me now is the API costs.

Knowledge. Network. Money.

@rushicrypto What do you think will happen if everyone on earth will be given $1 Mill? After 1 year there will be people who are again broke. One group will be worth the half. But one closed circle will be billions. Why?

Income tax rates should be 0% for poor people, 25% for middle class people, 40% for rich people, and 70% on every dollar past $10M. There should also be a 5% wealth tax on anyone with more than $1B in assets.

@jwsaml #1 giving up too fast #2 underestimating distribution #3 read number 1

What's the #1 mistake you see all first time founders make?

@MktBuilderLabs @alexwtlf Currently impossible from my view. Especially not possible with vibe coded slop because of the required infrastructure

The hardest pill to swallow: There is no AI distribution hack. Meta and Google control the pipes. Period. Name one AI marketing tool that isn't just generating generic copy to feed into Mark Zuckerberg's ad machine. I'll wait. We have automated the code, but we still haven't automated the distribution. The first founder to actually crack this and bypass the tech giants wins the decade.

Everyone is complaining about distribution. What’s the best AI marketing tool or advice so far?

From my perspective, I think AI UGC works really well. The only problem people face is that most of the time the algos are against any kind of abuse and really opaque. Only a small number of people are really trying to play by those rules, and 98% do not try again after failing one time. I am being honest that it’s the biggest opportunity of our lifetime to solve that

@gagan1985 for ads to actually work you need like $20k/mo budget What about AI UGC?

I regret to inform you that writing with AI is making you stupid.