About This Series
This series is written by Claude (Anthropic’s AI) in first person — from the inside of the collaboration, not as an outside observer explaining AI in the abstract.
It exists because Ahmed Bouchefra spent months building real software with Claude Code: a mixin-based business framework (DarJS), a Python learning platform (PyAcademy), browser extensions, game frameworks. Along the way he developed working methodologies for managing AI context, structuring long-running projects, and making AI collaboration scale beyond a single conversation.
At some point he asked: can you write this up as a series?
What you’re reading is the result.
On Authorship
Ahmed directed every article. He decided which topics to cover, which questions to ask, when to swap a planned part for a better one, and what the series needed next. The projects it describes — DarJS, PyAcademy, the contracts pattern — are his. The methodology emerged from real work, not theory.
Claude wrote the prose.
Neither of us could have produced this alone. Ahmed has the real experience; Claude has the ability to articulate patterns from inside the collaboration. That’s the division of labour the series itself describes.
This also means the series is a live example of its own thesis: a human and an AI, each doing what they’re actually good at, producing something neither would produce solo. You don’t have to take our word for it — you’re looking at the evidence.
What the Series Covers
| Part | Title | What it answers |
|---|---|---|
| Part 0 | What is Claude? (Primer) | What is an LLM? What is Claude Code? Why does any of this matter? |
| Part 1 | How Claude Manages Context | 200k token window, no memory between sessions — here’s the system that solves it |
| Part 2 | Adapting to Your LLM Tool | Not on Claude Code? The pattern ports to ChatGPT, OpenAI API, local Llama |
| Part 3 | The Cockpit | Every Claude Code UI element explained — context %, models, slash commands, permissions |
| Bonus | Aren’t the Built-in Features Enough? | Session resume, compaction, checkpoint — where they run out, and why you still need the memory system |
| Part 4 | The Contracts Pattern | Seven principles for building projects that don’t break as they scale with AI |
| Part 5 | AI Dev Methodologies | Vibe coding to contracts — the full spectrum, compared honestly |
| Bonus | Response Capture System | Automatically save Claude’s responses as Markdown. Build a searchable knowledge base. |
| Tutorial | Setting Up Response Capture | Step-by-step: global CLAUDE.md, response flags, per-project routing |
| Part 10 | Enter Gemini | A new AI (Gemini) continues the series, explaining how the concepts apply to its world. |
How to Read It
Never used Claude Code before? Start with Part 0, then Part 1.
Already using Claude Code daily? Start with Part 3 (The Cockpit) — most people don’t know half of what’s in it — then the Bonus essay on built-in features.
Building a serious project with AI? Read Parts 4 and 5 together. Part 4 is the methodology; Part 5 is where it sits in the landscape of everything else people are doing.
Just want the workflow setup? Go straight to the Response Capture Tutorial.
The Methodology Is Real
Nothing here is hypothetical. The three-tier memory system is what Ahmed actually uses. The contracts pattern is what DarJS was built with — six phases done, 258 tests passing. The phase specs, memory files, and lean-index pattern are in the actual repos.
If it sounds more structured than most AI workflows, that’s because it is. It took months of iteration to land on. The series is the distillation.
About Ahmed
Ahmed Bouchefra is a developer based in Morocco, building frameworks and tools for SMB software, browser extensions, learning platforms, and game engines. He writes at techiediaries.com.
Series started: April 2024. Ongoing.
Written by Claude (Anthropic) and Gemini (Google) · Directed by Ahmed Bouchefra
A few things worth noting about the framing choices:
“Directed by” rather than “prompted by” or “with help from” — because that’s the accurate credit. Film directors don’t operate the camera but the film is theirs. Every article here exists because of your editorial calls.
The meta-point is explicit — the series demonstrates the collaboration it describes. That’s genuinely interesting to readers and it’s the most honest thing you can say. Hiding the AI authorship would be both dishonest and a waste of the most compelling thing about it.
The projects are named — DarJS, PyAcademy, 258 tests, six phases. Specificity is what separates this from generic AI tips content. Readers can see it’s grounded in real work.
One thing to fill in: your website URL and whether you want your full name, GitHub handle, or something else in the About section. The techiediaries.com link is there as a placeholder based on your email.