How can someone just starting out earn money in the AI space run AI agents and local models, given that big companies have the resources?
I understand the concern about big model companies dominating, but there are two main types of improvements happening right now: hardware and software algorithmic improvements. Hardware advances mean larger server clusters and compute resources, which let companies train bigger models faster. Software improvements—quantization, inference hacks, and other algorithmic tricks—make models run faster and smaller. Because of these algorithmic gains, especially from teams that lack massive capital, we can run efficient models locally. For example, the Chinese models like GLM 5.2 are highly optimized and can be run on a consumer laptop: a 1‑bit quantized version runs on a MacBook with 256 GB of RAM. In the next six months we’ll likely be able to run comparable models on phones. So the fear of being locked out is overstated. The simplest way to get started is to download Ollama, a platform that lets you run open‑source models locally. You can pull models such as Ornith, GLM 5.2, or Kimmy K 2.7‑code, and many community members provide quantized versions that run on modest hardware. Even if a model is too large, you can pay a few dollars for a hosted service to help set it up, and you’ll be fine.
I’m not an expert in crypto, but models will fail a lot—about 70% of the time. The goal isn’t to prevent failures but to make them happen quickly so we can learn. Models are highly parallelizable; you can spin up dozens, test many approaches, discard what doesn’t work, and feed the results into the next model. By iterating fast, you eventually get a solution that works for complex, technical use cases.
I started with Harvard CS50, then taught myself computer science from the ground up, excluding languages and compilers, distributed systems, and operating systems. I completed nanotech to Tetris, building a computer from scratch. After that, I learned AI from scratch via a course that just listed introductory material, then I began training models—first a TensorFlow model called Stalan 2 that generated faces, then Gwarn's 'This Wu Does Not Exist' which I turned into an anime face generator, which I found hilarious and turned into a business called One Second Painting. I scaled that to about $3,500 per month by selling AI-generated art, but Midjourney's release destroyed that business. I had been selling NFTs, which gave me hands‑on AI experience. I then used that experience to freelance write while aiming for $10k/month, employing GPT‑3 for content. I also spent six months to a year locked in my room learning make.com and Zapier, building automations for my One Second Painting business and later for a second venture with a business partner. Eventually the money came from disproportionate time invested, leading to a compressed business model: get your first customer quickly, then worry about everything later. I realized most of the low‑level AI skills I learned aren’t used today, but the practical experience with automation platforms like make.com and Zapier, plus building info products, allowed me to scale to $100k/month through a blend of an AI automation agency and information products.
The feeling you describe is akrasia — knowing what to do but not doing it because the required activation energy feels too high. To overcome this, consider the two business models you mentioned. A SaaS model is highly scalable but requires product‑market fit and can take years to generate revenue, during which you burn runway with no income. An agency model, by contrast, lets you earn money quickly (e.g., $20k in the first month) but is hard to scale and lacks predictable recurring revenue. I recommend a blended approach: start with an agency to generate early cash flow and runway, use that revenue to fund your SaaS development, and then gradually shift from manual implementations to a SaaS product. This path takes longer to reach the same scalability as a pure SaaS, but it yields far more money in the early years and gives you proven business‑owner credibility, which makes raising funds or reinvesting easier.
AI will create ads, films, and music videos much faster and for far less money—no production or audio teams needed, as the AI handles scripting and auto‑fixes everything. The output comes from natural‑language prompting; those who know how to prompt the models will do best, though that skill will fade as models become better at prompting themselves. You can definitely earn $10k/month with this right now, maybe for a couple of years. I’m still a beginner in AI automation agencies, but I know some basics and have built a few automations. Your next step is to get in front of paying customers, articulate their needs and problems, and show how your solution helps. There’s no other path—it’s messy, like entering a gladiatorial arena with tigers; you have to win your first fight to earn praise. Start by signing up on Upwork or freelancer.com, treat yourself as a low‑paid API endpoint, gain experience and money, then move to cold email, ads, or other channels to build a sustainable method.