Skip to main content
Lesson 9 of 14 · Vibe (formerly Le Chat) Advanced: Deep Research, Canvas and More

Advanced Personalization

Lesson 9 of 14 Updated November 2, 2026 Intermediate
On this page
  1. Build your custom assistant
  2. The duo in action, by profession
  3. Global preferences and reusable templates
  4. When layers conflict
  5. Key takeaways

Update November 2, 2026 — Le Chat is now Vibe. As of May 28, 2026, Le Chat is now Vibe (still at chat.mistral.ai): a single agent for work and code, carrying over your conversations, settings, and subscriptions. The legacy Le Chat experience remains accessible in Chat mode; in Work mode, agents are now skills (invoked with @), memories are the knowledge base, and deep search is the Deep Research skill. Plans are now called Mistral Free, Pro, Team, and Enterprise. This lesson now uses the name Vibe; some menus described here may have changed name or location.

Build your custom assistant

Memories and Projects, discovered separately in previous lessons, only show their real value when combined: together, they produce an assistant that knows both your profession and the specific file you’re working on right now.

The distinction comes down to one sentence. Memories retain who you are, regardless of the current file: your role and industry, your communication style, your usual tools, your recurring constraints on budget, regulations, or team. Projects describe the context of a specific mission: reference documents, objectives, communication rules specific to the file, stakeholders involved. In a project, Vibe mobilizes both layers simultaneously—you get an interlocutor who knows you’re a consultant and that this specific client is in luxury goods.

The duo in action, by profession

A digital transformation consultant will enter their profession in Memories, their client size—“My clients are French mid-cap companies, 100 to 5,000 employees”—and their preference for structured responses with actionable recommendations. Each client then becomes a distinct project. For Client A, the library contains the initial audit, org chart, and roadmap: “The main contact is the CIO. Cloud migration is the priority. Approved budget: €200K.” For Client B, the marketing brief and competitive analysis, in a completely different register: “The client is in the luxury sector. High-end tone. The project concerns redesigning the digital customer experience.” The consulting style remains yours, but the vocabulary shifts from one project to another without you asking.

A developer follows the same logic. Their Memories carry the durable foundation—Python and TypeScript, clean code and design patterns, Agile team—their projects the current stack. The API project contains the OpenAPI specification and database schema: “FastAPI framework. PostgreSQL database. Always include types and docstrings.” The frontend project brings together the design system and exported Figma mockups: “React 19 + strict TypeScript. Tailwind CSS. Tests with Vitest.”

An HR manager at a 300-person company will enter the relevant industry labor agreements in memory and their obligation to comply with GDPR in all communications. Their recruitment project gathers job descriptions, salary grids, and internal processes: “Write job postings in inclusive language. Systematically mention social benefits.” Their training project contains the catalog, budget, and evaluations: “Training must be eligible for the company training budget where possible. Hybrid format preferred.” Two frameworks, one professional identity.

Global preferences and reusable templates

Beyond project-specific instructions, Vibe settings let you define global preferences: default output format (Markdown, plain text, or HTML), response language, level of detail, topics to handle with care. These settings apply everywhere.

For recurring tasks, build yourself instruction templates ready to paste into a new project. A “Competitive Analysis” template fits in one instruction: “Analyze each competitor according to these criteria: positioning, pricing, strengths, weaknesses, threats to us. Table format when possible. Sources must be cited.” An “SEO Writing” template will be just as operational: “Each article must include an H1 title with the main keyword, H2 subheadings every 300 words, a 155-character meta description, and internal linking to our existing pages.”

When layers conflict

Sometimes a global memory conflicts with a project instruction. The classic case: your memory says “Respond in French” and you open an international file that requires English. The rule is stable: project instructions take precedence over Memories. Nothing stops you from being explicit: “For this project, ignore my usual language preference and respond in English.”

Four pitfalls await those who personalize heavily. Accumulating useless memories blurs the profile instead of refining it: delete what you no longer need. Project instructions written months apart end up contradicting each other or your Memories, hence the interest in reviewing them. A change of role or constraints makes your memories false if you don’t update them. Finally, remain capable of providing explicit context outside your usual spaces: personalization is a tool, not a crutch.

The real-world test takes ten minutes. Enter five durable facts about your activity in your Memories, create a project around your current main project with its documents and instructions, then ask increasingly specific questions. If the answers bear both the mark of your profile and that of the file, the combination works.

Key takeaways

  • Memories defines your persistent identity, Projects defines the mission context
  • Combining both creates a truly personalized assistant
  • Prepare reusable instruction templates for your recurring tasks
  • Project instructions take precedence over Memories in case of conflict
  • Keep your Memories up to date and avoid information overload