Roadmap
Six phases, 12 things from this library, the people to follow at each stage, and the news topics worth watching while you work through it.
Every list on this site answers "what is there?". This page answers "what do I do first?". It is one route through the AI part of the library, laid out over about six months, with the people to follow and the news to watch at each stage.
The order matters more than the speed. If a phase takes you twice as long as it says, that is the phase working, not you failing — the timings assume a few hours a week around a job.
About six months at a few hours a week — the first useful week is week one.
0 of 6 steps done
Loading your saved progress…
Week 1
Opinions about AI are cheap and mostly secondhand. Spend the first week putting your own real work through two or three assistants — a job application, a plan, a spreadsheet you dread — and you will have your own evidence about what they do and do not do. All of these have a free tier; the free tier is all this phase needs.
By the end of it: You can say, from your own use, what these tools are good at and where they confidently make things up.
Worth knowing: Free tiers have daily limits and swap in smaller models when busy, so a weaker answer is not always the tool's ceiling. Never paste anything confidential into one.
OpenAI's conversational AI assistant for writing, research, brainstorming and support tasks. Free tier with message and model limits; paid plans lift them.
Why I recommend it: The single most useful free tool for brainstorming marketing copy and business ideas.
Anthropic's AI assistant. Strong at long writing, reading documents you paste in, analysis and coding help. Free tier with daily limits; paid plans lift them.
From the site: Claude is Anthropic
Why I recommend it: The one I reach for when the task is writing or thinking through a document. Check anything factual yourself — it can be confidently wrong.
Google's AI assistant, wired into Gmail, Docs and the rest of Workspace, so it can work on files you already have. Free tier; paid plans add the larger models.
Why I recommend it: Worth it mainly if your work already lives in Google Docs and Gmail — that integration is the real advantage.
Headlines are loading — see the news page for the full set.
Follow this topic on the news pageWeeks 2–10
This is the phase most people skip, and skipping it is why their AI knowledge stays at the level of tips. Elements of AI is free, non-technical and gives you the vocabulary. Harvard's CS50 AI course is the step up with real code. Microsoft's beginners' course is the shortest route to building one small thing yourself.
By the end of it: You know what training, a model and a token are, and you have built one small thing that calls a model.
Worth knowing: None of the three is a credential anyone hires on. They are the groundwork that makes the later phases possible.
Free, no-math introduction to artificial intelligence from the University of Helsinki.
Why I recommend it: The clearest starting point if AI still feels abstract. You will be able to talk about it accurately in an interview.
Harvard course covering graph search algorithms, optimization, machine learning, and natural language processing with hands-on Python projects.
Why I recommend it: The most rigorous free option here. Finish the projects — a CS50 project portfolio carries real weight with hiring managers.
21 structured lessons on prompt engineering and building generative AI applications, with practical exercises in Python and TypeScript.
Why I recommend it: The best free course for actually building something. Work one lesson at a time and keep the code you write — that is your proof of skill.
Headlines are loading — see the news page for the full set.
Follow this topic on the news pageWeeks 8–12, alongside the courses
Every model launch comes with numbers chosen by the company that made it. These three sites publish independent comparisons — human head-to-heads on LMArena, speed and cost on Artificial Analysis, hard benchmark results at Epoch. Reading them for a month teaches you to spot a claim resting on the maker's own test.
By the end of it: Given a launch announcement, you can find out within ten minutes whether the claim holds up independently.
Worth knowing: Leaderboards measure what they measure. Human preference votes reward answers that read well, which is not the same as answers that are right.
Head-to-head model comparisons voted on by the public: you see two anonymous answers to the same prompt and pick the better one, and the rankings come from those votes. Free.
From the site: Chat, compare, vote for the world's best AI models. Join the community shaping the public leaderboard for LLMs, image, and code models through real-world evaluation.
Why I recommend it: The closest thing to a fair fight between models on ordinary prompts, instead of marketing claims. Votes are taste as much as accuracy, so read it as popularity with a purpose.
Independent benchmarking of the major AI models on speed, price and quality, with the numbers side by side. Free to read.
From the site: Comparison and analysis of AI models and API hosting providers. Independent benchmarks across key performance metrics including quality, price, output speed & latency.
Why I recommend it: Where I check what a model actually costs per million words before believing a "cheap" claim.
Research-grade tracking of what AI models can do and how that has changed over time, with the data and methods published. Free.
From the site: Our hub for benchmark results, featuring the performance of leading AI models on challenging tasks. It includes results from benchmarks administered internally by Epoch AI as well as data collected from external sources. Explore trends in AI capabilities across time, by benchmark, or by model.
Why I recommend it: For the longer view rather than this week's launch — they show their working, which most leaderboards do not.
Headlines are loading — see the news page for the full set.
Follow this topic on the news pageMonths 3–4
Whatever your job title, the AI conversation at work will turn to risk, bias and rules, and the NIST framework is the free vocabulary US employers cite. Read it once, then write a two-page risk register for one AI system you actually use. The papers behind the disagreement about how serious all this is — and what to read in what order — are the deep dive, not this phase.
By the end of it: You have written a two-page risk register for one AI system you actually use, in the language governance people recognize.
Want to go deeper on AI safety and evaluation specifically? The full version of this phase — the papers in order, the frameworks and the tools you run yourself — lives in one place.
See the full AI Safety learning pathHeadlines are loading — see the news page for the full set.
Follow this topic on the news pageMonths 4–6
This is where the roadmap turns into something on a resume: you point a free scanner at a model, watch a jailbreak land, and write up what failed and how often. One written-up scan beats any certificate. The install commands, the tools in order and what to put in the report are all in the safety path.
By the end of it: A short report: what you probed, what failed, how often, and what you would change — mapped back to the NIST wording.
Want to go deeper on AI safety and evaluation specifically? The full version of this phase — the papers in order, the frameworks and the tools you run yourself — lives in one place.
See the full AI Safety learning pathHeadlines are loading — see the news page for the full set.
Follow this topic on the news pageOngoing, from month one
The field moves weekly, and trying to read everything is how people burn out and quit. Twenty minutes twice a week on one news topic and three or four people is enough to stay current. The structured free course is the thing to do last, when you have your own findings to bring to the discussions.
By the end of it: A habit you can keep: two sittings a week, one topic, a handful of people, and one course finished with a real project.
Worth knowing: Course cohorts are competitive and run to a schedule, so you may wait for an intake — the curriculum is published free either way.
A free structured course in AI alignment and AI governance — readings, exercises and facilitated cohorts. Self-paced version free to anyone.
From the site: Free online courses, grants, and intensive in-person programs from the leading talent accelerator for beneficial AI and societal resilience. Join 10,000+ alumni and start today.
Why I recommend it: The usual route in for people trying to move into safety work. The reading list alone is worth the visit even if you never join a cohort.
The open-access preprint server for physics, mathematics, computer science, and related fields — a primary source for cutting-edge AI and machine-learning research papers.
Why I recommend it: The best place to read AI research before it hits journals or the press; search by tag or author to follow a specific line of work.
Open-source AI community and platform hosting machine-learning models, datasets and tools. Free to browse, download and run models; paid plans only cover hosted compute.
Headlines are loading — see the news page for the full set.
Follow this topic on the news pageThis page gets you oriented across all of AI. When you want to go deep on one thing, the AI safety and evaluation path takes the papers, the frameworks and the tools in order, with something to show at each step. Everyone named above is on the people pages, and the wider collection is under Tech & Ethics.
