Get AI summaries of any video or article — Sign up free
Projects—12 Days of OpenAI: Day 7 thumbnail

Projects—12 Days of OpenAI: Day 7

OpenAI·
5 min read

Based on OpenAI's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.

TL;DR

Projects organizes ChatGPT conversations around a shared goal using project-scoped files and custom instructions.

Briefing

Projects is rolling out inside ChatGPT as a new way to organize conversations around a shared goal—complete with file uploads, custom instructions, and tools like search and Canvas working within the same project context. The practical payoff is that users can stop treating every question as a fresh thread and instead build a reusable workspace where related chats, documents, and outputs stay connected.

The rollout comes with a clear set of capabilities. Creating a project adds a title and optional color in the left sidebar, then lets users attach files and instructions that apply to the project’s conversations. Users can also bring existing chats into a project, using conversation search to find relevant past threads and then add them to the same workspace. In effect, Projects functions like “smart folders” for ChatGPT activity—designed to keep context, references, and outputs from scattering across the account.

A live demonstration showed how Projects can support both lightweight organization and deeper, data-driven workflows. In a Secret Santa example, uploaded survey results become project files, and project instructions guide ChatGPT to read that spreadsheet, randomly assign givers to recipients, and output a table with giver, recipient, and gift ideas. The workflow also included validation: the organizer referenced Python code used to shuffle assignments, reinforcing that the results weren’t arbitrary. From there, the project context enabled follow-up questions—such as what a specific person wants and who is assigned to give it—without re-uploading or re-explaining the underlying data.

Projects also ties into web search and Canvas. After assignments are generated, ChatGPT can search the web for gift ideas (for example, festive hats) and then draft communications. Canvas support lets users generate and iteratively refine longer text like emails; the demo showed an initial draft that became more specific after feedback, with Canvas enabling a faster edit loop than starting from scratch.

Beyond holiday planning, the transcript highlights personal and technical use cases. A “Home Maintenance” project pairs task logs with appliance documentation so questions like whether a refrigerator water filter needs replacing can be answered using the user’s own notes and history (including timing such as “overdue since March”). Another example used Projects to update a personal website template built with Astro and Tailwind-style styling: the user uploaded resume and social links, provided custom instructions about Astro formatting, and then used Canvas to modify the baseline template. The workflow included diff-style edits so changes were visible, and follow-up steps refined specific sections like testimonials.

Projects is launching for Plus, Pro, and Teams users starting immediately, with free users to follow soon. Enterprise and EDU availability is slated for early next year, with a gradual rollout during the day to manage load and stability.

Cornell Notes

Projects brings a structured workspace to ChatGPT, letting users group related conversations under a single goal with shared context. Each project can include uploaded files, custom instructions, and project-scoped use of tools like search and Canvas. The Secret Santa demo illustrated how spreadsheet data can drive assignments and follow-up questions without re-entering details. Other examples showed practical personal use (Home Maintenance logs and appliance notes) and technical use (editing an Astro-based website template with diff-style Canvas changes). This matters because it turns scattered chat history into reusable, referenceable workspaces.

How does a user create and customize a Projects workspace in ChatGPT?

A user creates a project from the left sidebar by clicking “plus,” then sets a project title and can choose a color for the sidebar. Projects can include uploaded files and custom instructions that apply across conversations inside that project. Users can also start new chats within the project, keeping the project’s context attached to the conversation.

What makes Projects different from simply starting a new chat each time?

Projects keeps related work connected. Instead of re-uploading documents or re-stating instructions, a project can store files (like a spreadsheet of survey results) and instructions (like event rules or formatting requirements). It also supports adding existing chats into the same project using conversation search, so past work can be reused as part of the same organized workspace.

How did the Secret Santa example use project files and instructions to automate a workflow?

The demo uploaded Secret Santa survey results as a project file and included instructions to assign each person to another person and generate a table with giver, recipient, and gift ideas. The model pulled from the spreadsheet to produce assignments (e.g., identifying who gives to whom and what the recipient wants). It also referenced validation via Python shuffle code to support the randomness claim, then enabled follow-up questions using the same project context.

What roles do web search and Canvas play inside Projects?

Web search helps generate ideas using external information within the project context—for example, searching for festive hats after assignments are made. Canvas supports drafting and iterative refinement of longer content like emails; the demo showed an initial email draft that became more specific after feedback, with Canvas enabling a faster edit loop.

How can Projects support personal knowledge management, not just one-off Q&A?

The Home Maintenance project combined a maintenance log with written documentation for appliances and a smart home setup. That pairing allowed questions like whether to replace a refrigerator water filter to be answered using the user’s own history (including timing such as being overdue since March). It also supported “how do I do it?” questions by pulling steps from the uploaded fridge notes.

How can Projects help with programming or technical editing tasks?

The transcript’s website example used Projects to tailor ChatGPT to an Astro template workflow. The user uploaded personal content (resume, social links, testimonials) and added custom instructions emphasizing Astro formatting and a quirky tone. Canvas then modified the baseline template, showing diff-style changes, and follow-up prompts refined specific sections like testimonials without rewriting the entire page.

Review Questions

  1. In what ways can a project’s files and custom instructions reduce repeated work compared with normal chat threads?
  2. Describe how the Secret Santa workflow used uploaded data to generate assignments and then supported follow-up questions.
  3. What advantages did Canvas provide during the Astro website template editing process, compared with editing large blocks of text?

Key Points

  1. 1

    Projects organizes ChatGPT conversations around a shared goal using project-scoped files and custom instructions.

  2. 2

    Users can create projects from the left sidebar, set titles and colors, and start chats that inherit the project context.

  3. 3

    Projects supports adding existing chats into a project using conversation search, turning past work into reusable “smart folders.”

  4. 4

    Uploaded spreadsheets and instructions can drive structured outputs like assignment tables and enable follow-up questions using the same data.

  5. 5

    Web search and Canvas work within Projects to generate ideas and draft longer content that can be iteratively refined.

  6. 6

    Projects can function as personal knowledge management by pairing logs with documentation for appliances and home systems.

  7. 7

    Projects is rolling out first to Plus, Pro, and Teams, with free users soon and Enterprise/EDU in early next year.

Highlights

Projects turns related conversations into a reusable workspace by binding uploaded files and custom instructions to every chat inside the project.
The Secret Santa demo showed spreadsheet-driven automation: assignments were generated from uploaded survey results and then validated with referenced shuffle code.
Canvas inside Projects enables iterative drafting—like emails—where feedback produces more specific outputs without restarting from scratch.
A Home Maintenance project answered practical questions by pulling from a maintenance log and appliance notes, including timing details like being overdue since March.
Technical editing was demonstrated with an Astro-based website template, where Canvas produced diff-style changes guided by custom instructions about formatting.

Topics

Mentioned