To say "I wanna work with AI" is as broad as saying "I wanna work with something on the internet."

The key difference between "internet" and "AI" statements is the speed at which things are happening right now.

If we look at the internet and talk about protocols: we talked about IP, and we had IPv4 for a long time. We were close to running out of IP addresses, then someone invented IPv6. It took a while to get there, but it's actually quite a revolution to move from one IP stack to another.

Then comes the learning path. You need to learn about the new TCP/IP version and how you set up networks for that, and you need to understand which vendor has the right firmware and stuff like that. You have the time to figure it out, get it working, and then you're in the loop.

Now let's talk about AI. You can say "I wanna work with AI," "I wanna do automations with AI," "I wanna do web coding," "I wanna do troubleshooting." That's okay — you don't need deep knowledge in all those areas to start, and you'll pick it up.

The key difference is speed: things in AI evolve much faster than they did on the internet. Knowledge can be three weeks old and already not worth much. That's how fast AI is developing right now. It's an exciting time, but also a hard time for your brain — you need to keep up, otherwise you'll be left behind. You might say "I can't take three weeks off because what I just learned could be outdated in three weeks," and then you'd have to start over.

This part above: I only asked CoClaw to spellcheck and remove double words, and to ask me if what I sent was done with the (handy) speech-to-text free tool.

I asked my CoClaw :) https://jp-my-blog.vercel.app/blog/coclaw-your-ai-powered-telegram-terminal-in-your-pocket-

"Can you make a list of what an AI agent builder needs in terms of skills?"

It gave me a long list with explanations, then I responded: "Just make a list in headline format." So the issue here is: I did that, CoClaw did it, and I can't find anything I'd call wrong — but to be honest I don't know what some of the points mean. The key point is I could have made something similar, but getting into writing mode and double-checking that I had all the headlines would have taken me about 100× longer.

Headline list (with links)

• Understanding Large Language Models (LLMs) — https://huggingface.co/docs/transformers
• Prompt Engineering & Instruction Design — https://platform.openai.com/docs/guides/prompting
• Embeddings & Semantic Search — https://www.sbert.net/ and https://platform.openai.com/docs/guides/embeddings
• Retrieval‑Augmented Generation (RAG) — https://langchain.readthedocs.io/en/latest/use_cases/rag.html
• Agent Planning & Hierarchical Reasoning — https://python.langchain.com/en/latest/modules/agents/
• Tooling & Safe Tool Orchestration — https://python.langchain.com/en/latest/modules/agents/tools.html
• Memory Design: Short‑term vs Long‑term — https://python.langchain.com/en/latest/modules/memory/overview.html
• Multi‑Agent Coordination & Roles — https://en.wikipedia.org/wiki/Multi-agent_system
• Reinforcement Learning & RLHF Basics — https://arxiv.org/abs/1706.03741 (Deep RL from human preferences)
• API Design & Backend Engineering — https://restfulapi.net/
• Workflow Orchestration & Scheduling — https://airflow.apache.org/ and https://temporal.io/
• Data Collection, Annotation & Labeling — https://labelbox.com/ (example tooling)
• Synthetic Data Generation & Augmentation — https://en.wikipedia.org/wiki/Synthetic_data
• Model Serving & Deployment Strategies — https://docs.bentoml.org/ and https://developer.nvidia.com/nvidia-triton-inference-server
• Infrastructure, Scaling & Cost Optimization — https://kubernetes.io/ and https://www.ray.io/
• Vector Databases & Retrieval Infrastructure — https://www.pinecone.io/, https://milvus.io/, https://weaviate.io/
• Observability, Logging & Evaluation Metrics — https://prometheus.io/, https://grafana.com/, https://wandb.ai/
• Safety, Privacy & Governance Controls — https://gdpr.eu/
• Security, Auth & Secrets Management — https://www.vaultproject.io/ and https://owasp.org/
• Human‑in‑the‑Loop Integration — (see Labelbox / active learning patterns) https://labelbox.com/
• Experimentation, A/B Testing & User Studies — https://mlflow.org/ and https://wandb.ai/
• Research Literacy & Experimental Design — (general research resources; e.g., arXiv) https://arxiv.org/
• Cross‑Functional Communication & Documentation — (docs & playbooks; e.g., docs-as-code approaches) https://www.writethedocs.org/