About the name

Yukti (Sanskrit: युक्ति) means skill, logic, ingenuity — the applied reasoning that turns knowledge into something useful. yuktics is yukti applied to computer science: the curriculum for students who want to do CS, not just study it.

What this is

A complete computer-science curriculum, rebuilt for students starting now. It covers the same territory as a traditional CS degree — programming, theory, systems, building, career — plus the things that have to be added in 2026: AI literacy, modern tooling, the new way of learning. Eight tracks, 44 modules, all opinionated.

Think of it as the replacement for grinding GeeksForGeeks five years ago. Same job (give a CS student a clear path from "I'm new" to "I'm employable"), updated assumptions (AI is a tool you use to learn, the modern stack is web + Python + AI, theory is still essential but you build it instead of memorizing it).

Who this is for

High schoolers wondering if CS is for them.
CS freshmen and sophomores who suspect their syllabus alone won't be enough.
Late-degree students compressing toward internships and interviews.
Career switchers / bootcamp grads who want CS depth, not just a framework certificate.
Self-taught devs filling the theory holes that didn't matter until they did.

Who this is not for

People looking for a no-code AI path.
People preparing for a research PhD specifically — this is for builders.
People who want a list of tutorials. This is a curriculum, not a Notion dump.

How the curriculum works

Eight tracks (T0–T7) hold modules. Each module is a 4–14 hour unit of work with the same five-part shape:

Outline. What you'll be able to do at the end. Specific.
Stack. Exact libraries, models, APIs, CLI tools.
Resources. Papers, repos, posts — ranked, not piled.
Build. A concrete project with milestones. The point of the module.
Checkpoints. Five questions you should be able to answer cold.

Modules name their prerequisites explicitly so you can skip around. Within a track, ordering is intentional but not enforced. The roadmap on the homepage is the index; the four learning paths are pre-built routes for common starting points.

What to do today

If you do nothing else this week, do these three things, in this order:

Pick the learning path that sounds most like you on the homepage. Not the most ambitious one — the one you'll actually finish.
Read module 00.3 — Using AI as your tutor. It is short. It is also the single highest-leverage 90 minutes you will spend in this curriculum, because it will change how every subsequent module compounds.
Set up your dev environment below. Don't spend more than 90 minutes on this. Get something working, fix the rest as it bites you.

Dev environment, minimum viable

The full curriculum needs the following on a Mac, Linux, or WSL machine.

# 1. A Unix shell
#    macOS: Terminal.app or iTerm2 are fine.
#    Windows: install WSL2, then open Ubuntu.
#    Linux: you're already there.

# 2. Homebrew (mac) or apt (linux/wsl) — your package manager
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

# 3. Core CLI tools
brew install git gh tmux ripgrep jq fd bat

# 4. Python via uv (fastest, cleanest manager in 2026)
curl -LsSf https://astral.sh/uv/install.sh | sh
uv python install 3.12

# 5. Node via fnm (clean per-project versions)
brew install fnm
fnm install --lts

# 6. An editor — VS Code or Cursor for the GUI side
brew install --cask cursor   # or: brew install --cask visual-studio-code

# 7. Claude Code (or equivalent) — covered in module 00.3
npm install -g @anthropic-ai/claude-code

# 8. API keys for AI tools (you'll need at least one)
# Add to ~/.zshrc or ~/.bashrc:
export ANTHROPIC_API_KEY=sk-ant-...
export OPENAI_API_KEY=...     # optional but useful for compare-runs

Hardware reality

You don't need an H100. Concretely:

T0 / T1 / most of T2: any laptop made in the last five years.
T3 (web/backend): same. Cloud free tiers are enough for shipping.
T4 — most of it: Mac (M-series, 16GB) or any laptop. Free Colab fills the gaps.
T4 — heavy fine-tuning: rent a single A10/L4 from Modal/RunPod for $0.50–$1/hr.
T4 — pretraining a frontier model: not in this curriculum. That's not a student exercise.

How to actually learn from this

Open a module. Read the outline and stack. Then do not read the body straight through. Skim it once for the shape. Open the build project and start it. Come back to the body when a step gets stuck. Read the resources only when you hit specific questions. Finish the build. Then run yourself through the checkpoints out loud, alone, in plain language. If a checkpoint wobbles, that's the section in the body you skipped past too fast.

The curriculum is short on prose for a reason. The thing that turns a module into knowledge is the project, not the page.

About the source

Drafts publish as they survive their own re-read. Ready modules have been used at least once end-to-end. The full curriculum is open source — if you find an error, a sharper resource, or a clearer way to phrase a checkpoint, open a PR. The repo is the book.