A shift you can't ignore

If you're even slightly in the loop, you can see what's happening. The owners of the big neural networks are fighting for clients. Token prices are artificially low right now. Open-source models (Llama, Mistral, DeepSeek) are squeezing from above. The market will split: cheap models for the masses, expensive ones for business. But either way, the barrier to entry in development has already dropped and will keep dropping. Picture the caliber of people who'll be left without work. These aren't dumb folks. Once they figure out neural nets — and they will — competition in development will get brutal. Yes, mastering agents takes months, I went through it myself. But those months will pass for everyone. It's already all about sales Code is becoming a commodity. Building a product is no longer the feat. The feat is getting it to the customer. And this isn't "in a year." This is the situation right now. Whoever has distribution has a business. Whoever just has code has a hobby. Organic + paid ads Native tools — Meta Ads, Google Ads — are expensive. But they're controllable. Betting everything on viral content is a lottery, even at industrial scale. The working strategy is a combination. Organic via Instagram, TikTok, YouTube — for reach. Paid ads — for conversion. The competition will come down to whose approach to content generation and whose promotion bots turn out to be more effective. Bottom line Don't focus on building the product. Start with how, what, and where you're going to sell — and assemble the product around that. The logistics have shifted. We'll come back to this topic. I'm dying to hear your ideas. The winner now isn't the one who writes better code. It's the one who's faster at: — assembling a system — and figuring out how to sell it Code will level out. Sales won't.

Claude Code in the field: where to look for the entry point Video generation — the window is closing After spending several months on video generators, I get it: the field is raw, more like a random generator. The only thing that actually works is ComfyUI. Installs locally, runs on rented GPUs, complex workflow. If you set your mind to it, you can teach Claude Code to work with ready-made templates and crank out videos automatically. That's the core for any system: marketing, account growth, ads. But: TikTok and Instagram algorithms are already learning to detect AI content and pushing it down in the feed. The window of opportunity is narrowing. If you're going to do it, do it fast. Gray zone There are niches where the main promotion services refuse to work: weapons, casinos, US dispensaries, webcam. Lower competition, demand exists, willingness to pay too. For a beginner, this can be a decent entry point. But don't build a long-term strategy on it — legal risks, payment processor blocks, account bans. Make money and move into a clean niche — that's the playbook. Poker and neural nets My favorite topic — training a neural net to play online poker. Solvers like PioSolver already solve poker mathematically, but for PLO they're significantly weaker than for hold'em. There's room here. The main question isn't "can you train it" — it's how to use it undetected. Poker rooms actively hunt bots. This isn't a product to sell — it's a personal weapon. I'll come back to this topic. Waiting for your ideas All of this is the directions I see right now. I'm curious to hear your thoughts: what do you see, where do you sense opportunities. I'm not looking for "an idea." I look where the money already is and where there are no decent solutions yet. If you see places like that — write to me.