DROPGoogle launches Gemini 3.8 Flash TTS and Flash-Lite TTS///DROPGemini Omni brings 1080p generation to Google Vids///BREAKClaude Opus 5.5 is here: 1M context, $4 input / $20 output///BREAKOpenAI adds GPT-6 Sol and Luna to the GPT-6 family///PAPEROpus 5.5's system card documents gains—and evaluation limits///BREAKGrok 4.7 launches at $2 input / $6 output per million tokens///DROPGrok Voice Transcribe 2.0 targets difficult real-world audio///TRENDGoogle steers new Gemini API projects to 3.8 Flash or 3.5 Flash-Lite///DROPGoogle launches Gemini 3.8 Flash TTS and Flash-Lite TTS///DROPGemini Omni brings 1080p generation to Google Vids///BREAKClaude Opus 5.5 is here: 1M context, $4 input / $20 output///BREAKOpenAI adds GPT-6 Sol and Luna to the GPT-6 family///PAPEROpus 5.5's system card documents gains—and evaluation limits///BREAKGrok 4.7 launches at $2 input / $6 output per million tokens///DROPGrok Voice Transcribe 2.0 targets difficult real-world audio///TRENDGoogle steers new Gemini API projects to 3.8 Flash or 3.5 Flash-Lite///
DARC.DEV

[ 01 — Start here · 6 lessons · ~3h with exercises ]

AI Foundations

Build the mental models behind useful AI: tokens, context, sources, capability boundaries and a defensible model choice. Finish with a small evidence-backed comparison.

← All learning paths

Your progress

0/6

Mark completion after doing the exercise. The checklist records your self-assessment, not a certificate. Stored for this browser only.

Module 01

Understand the system

Separate what the model knows from what the application supplies.

~25 min · read + exercise

Tokens, context and tools

Explain the four layers of an LLM application.

Exercise

Draw the model, context, tool and policy boundaries for a calendar assistant.

Done when: You can identify which layer prevents an unauthorized calendar write.

Read the lesson

~25 min · read + exercise

Read the current model map

Find exact IDs, availability and price caveats.

Exercise

Pick two current models and record their API IDs, modalities and rate conditions.

Done when: Every entry has a primary source and a checked date.

Read the lesson

Module 02

Learn with evidence

Turn research into inspectable claims, not a polished summary.

~35 min · read + exercise

Verify a source-backed answer

Distinguish a citation from supporting evidence.

Exercise

Audit five claims in an AI answer and record whether each source supports them.

Done when: Unsupported and ambiguous claims are explicitly labeled.

Read the lesson

~30 min · read + exercise

Build a research notebook

Curate a corpus and preserve disagreement.

Exercise

Create a brief from five sources with findings, disagreements and open questions.

Done when: You have opened every source behind a consequential finding.

Read the lesson

Module 03

Make a useful decision

Select for a job and evaluate without a universal leaderboard.

~30 min · read + exercise

Choose by the task

Separate hard constraints from preferences.

Exercise

Create a decision matrix for one workflow you actually use.

Done when: A model that fails a hard requirement is excluded even if its public score is high.

Read the lesson

~35 min · read + exercise

Design five acceptance cases

Write checks that a human can understand.

Exercise

Create three normal cases, one missing-evidence case and one denied-action case.

Done when: Each case has an observable expected outcome, not 'seems good'.

Read the lesson