A French AI account's repost of Boris Cherny's Claude Code talk reached 3.6 million views and 53,000 bookmarks. The original is Anthropic's 2025 "Mastering Claude Code in 30 minutes." Here are the actual techniques, what changed since, and how non-developers can apply them to AI employees.
On October 5, 2026, the French automation account Deep Genius AI (@Deep_GeniusAi) posted an 18-minute clip of Boris Cherny, the engineer who created Claude Code, explaining how to work with it. The post (in French) says the engineer "just released" a 28-minute video on writing prompts that work. In fact the clip comes from "Mastering Claude Code in 30 minutes", a talk Cherny gave at Anthropic's Code with Claude event and Anthropic published on YouTube on May 22, 2025. The ideas have held up, and they apply well beyond coding.
L’ingénieur qui a construit Claude Code vient de publier une vidéo de 28 minutes sur la façon d’écrire des prompts qui fonctionnent vraiment.
— || Deep Genius AI || AUTOMATION || AI || (@Deep_GeniusAi) October 5, 2026
J’ai vu des formations à 300 $ qui n’abordent même pas ce qu’il montre dans les 10 premières minutes.
( La suite de la vidéo est dans les commentaires. )
Fichiers CLAUDE.md, raccourcis mémoire, sessions parallèles, techniques de prompting…
Tout dans une seule vidéo, et entièrement gratuitement.
Que tu sois développeur, débutant ou que tu utilises Claude depuis des mois, ça peut vraiment t’être utile.
Le genre de thread que tu vas regretter de ne pas avoir mis en signet. 🔖
Post credit: video and post by || Deep Genius AI || AUTOMATION || AI || (@Deep_GeniusAi) on X, published October 5, 2026 (3.6 million views and 16,600 likes when we wrote this). View the post on X. All rights to the video belong to its creator; we embed it with X's standard embed and add our own commentary.
In English, the post says it has seen $300 courses that don't cover what Cherny shows in the first 10 minutes, and lists "CLAUDE.md files, memory shortcuts, parallel sessions, prompting techniques," all free in one video.
The speaker is Boris Cherny, who created Claude Code at Anthropic. The full talk, "Mastering Claude Code in 30 minutes", runs about 28 minutes and has more than 1.6 million views on Anthropic's YouTube channel. The slides in the X clip match the talk: "Optimize your setup," "Use Claude Code to answer questions about your codebase," "Give Claude more context" and "Share with your team." If you want to learn from it, watch the original. It has the full context, and the creator gets the credit.
Most of the talk is less about clever wording and more about giving Claude the right context, tools and checks. Here is what the slides cover, matched to Anthropic's current Claude Code best practices:
| Technique | What the talk shows | What Anthropic's docs say now |
|---|---|---|
| Start with questions | Ask Claude about the codebase before changing it, e.g. "What did I ship last week?" | Ask Claude the questions you would ask a senior engineer. It is an effective onboarding workflow. |
| CLAUDE.md context files | Files at company, personal, project and local level that load into every session | CLAUDE.md is read at the start of every conversation. Keep it short and prune it. |
| Memory shortcut | Type # to save an instruction to memory | See the memory docs for current ways to edit CLAUDE.md |
| Explore, plan, then code | "Figure out the root cause... propose a few fixes. Let me choose an approach before you code." | Use plan mode to separate exploration from execution |
| Verification | Write tests first, or have Claude screenshot its result and iterate until it matches a mock | "Give Claude a check it can run: tests, a build, a screenshot to compare." |
| Parallel agents | "Use 3 parallel agents to brainstorm ideas" | Run multiple sessions in parallel, for example a Writer/Reviewer pair |
| Plug in team tools | Tell Claude about your CLI tools ("use -h to learn it") and MCP servers | CLI tools are the most context-efficient way to reach external services |
| Share setup with the team | Check memory, commands, permissions and MCP config into the project | Check CLAUDE.md into git so your team can contribute |
The thread running through all of it: Claude can't read your mind, and it can't tell when it's done unless you give it a way to check. Cherny has kept to this since. In a January 2026 write-up, InfoQ reported that he runs 10 to 15 Claude sessions at once, starts pull requests in plan mode, and has Claude test every change it lands.
The principles hold up, but some details in the clip are out of date. The talk's custom slash commands in .claude/commands now sit alongside skills, folders with a SKILL.md that Claude loads only when relevant. Anthropic's docs now also cover plan mode, subagents for research, /clear between unrelated tasks, and a separate reviewer before you call work done. If a command in the video doesn't work for you, check the best practices page rather than assuming the idea is wrong.
One more correction: the X post calls it a video on writing prompts. It is really a talk on setting up the work, with prompts as one part.
The framing did most of the work. "The engineer who built Claude Code" gives authority. "$300 courses that don't cover what he shows" gives a bargain. "You'll regret not bookmarking" pushes saving over reading, which is why bookmarks (about 53,000) beat likes more than three to one. Posting the video natively, with "the rest in the comments," kept people on X.
It also landed at a good moment. Claude Code has moved far beyond its original audience, and many new users are looking for a short, credible guide. A free talk from its creator is exactly that, even 16 months later.
You don't need to write code to use the same ideas. Whether you run ChatGPT, Claude or an AI employee that answers your phone, the same four habits apply:
When the AI makes a mistake, add one line to the context file so it doesn't happen again. Over months, that file becomes your playbook. Our guides on automating business processes and AI employees vs virtual assistants go further.
What isn't relevant yet: running 10 parallel AI sessions. That suits engineers working on large codebases. Most small businesses get more from one well-briefed agent on one workflow than from many half-briefed ones.
Dooza is an AI-native company that builds AI products and services for small businesses, from the Dooza Workforce app to the Dooza Agents platform. Every product starts with a refundable pilot: 100% refund within 14 days.
Dooza applies these principles for you. Each AI employee in Dooza Workforce (Maily for email, Somi for social media, Ranky for SEO and AI visibility, Stan for lead generation, Linda for legal documents, and Rachel for phone calls) works from your business context and asks for approval on anything sensitive. Dooza Agents are custom agents that Dooza engineers build and maintain, including the instructions, context files and checks that keep them reliable.
Good first workflows are an AI receptionist, AI customer support, and workflow automation for repetitive admin. If you are comparing options, read Claude Cowork vs Dooza. Pricing depends on the product and is on our pricing page.
Book a free 30-minute call and a Dooza engineer will scope a pilot around one workflow, including the context and checks it needs. Start with a refundable pilot: 100% refund within 14 days. Book a free pilot call.
Boris Cherny, the creator of Claude Code. The clip comes from his talk "Mastering Claude Code in 30 minutes," published on Anthropic's YouTube channel on May 22, 2025. The viral X post by @Deep_GeniusAi is a repost.
No. The X post says it was "just released," but the talk was published in May 2025. Most principles still apply, though some commands have changed, so check Anthropic's current Claude Code docs.
A plain text file Claude Code reads at the start of every session. It holds commands, style rules and project conventions Claude cannot work out on its own. Anthropic recommends keeping it short and pruning it regularly.
Give Claude a way to verify its work, such as tests, a build or a screenshot to compare. Anthropic's best practices say this is the difference between a session you watch and one you can walk away from.
Yes. Write a short context file about your business, give specific instructions, ask for a plan before the AI acts, and set a check such as approvals or weekly reviews. The same habits make AI employees reliable.
Dooza Workforce employees and custom Dooza Agents work from your business context, follow clear instructions and ask for approval on sensitive actions. Every product starts with a refundable pilot: 100% refund within 14 days.
Automate your business with AI employees that work 24/7. Start with a refundable pilot: 100% refund within 14 days.
Anthropic's Claude Opus 5.5 is the first model in the Claude 5.5 family. It performs at roughly the level of Claude Fable 5.1 on most work, costs 40% less to run than Opus 5, and generates output more than 30% faster. Here is what changed, how to choose between Opus 5.5, Fable 5.1 and Sonnet 5.5, and what it means for businesses running AI agents.
Anthropic's Claude Sonnet 5.5 is the second model in the Claude 5.5 family. It keeps Sonnet 5's token prices, runs more than 30% faster, and Anthropic says it costs up to 30% less per task because it needs fewer tokens. Here is how it compares with Opus 5.5, where it fits best, and what it means for small businesses.
Start with a refundable pilot — 100% refund within 14 days. A Dooza engineer scopes it with you on a free 30-minute call. Pricing depends on the product; see pricing.
Refundable pilot · 100% refund within 14 days · No contracts