Bring an idea or the work you already have. These full briefs ask Astra to research, build and test a useful result you can put in front of real people.
5 full project briefsStart with what you haveChecks included
Choose Astra when it is available. These jobs need the relevant browsing, file and building tools; a local Codex project is useful for working previews. Have the task confirm what it can access and explain any step it cannot complete.
01 / Find a problem people already pay to fix
Start with customer calls, or an area you would like to build a business in. Let AI find the opportunity and turn it into something you can show people.
WHAT TO GIVE IT
Your customer-call transcripts, if you have them. Starting fresh? Give it an industry, audience or problem area and ask it to use public reviews.
Find a recurring customer problem worth building an offer around. Use my attached customer calls if supplied. If I am starting fresh, research public reviews and discussions in the industry or audience I name. Find the useful patterns yourself.
Look for frustrations, current workarounds and evidence that people spend money trying to solve the problem. Check competing offers and their current public pricing. Keep source links, dates and short supporting quotes. Distinguish customer reports from marketing claims; do not invent sales figures or treat an advertised price as proof of demand.
Choose the strongest opportunity and explain why, including contrary evidence and what is still uncertain. Turn it into a specific offer: intended customer, useful outcome, inclusions, delivery approach, scope boundaries and a proposed price or pricing questions for my review.
Build a complete sales-page draft around that offer, using an editable page or document. Keep unverified proof and details clearly marked. Check that the page matches the researched need and promised scope, and test its layout on mobile if you build a webpage.
Show the research, offer and finished draft. Do not publish, contact customers or take payments.
THE RESULT TO ASK FOR
An offer idea supported by research, proposed inclusions and a complete sales-page draft.
QUICK REVIEW
Does the research support the problem and proposed offer, with links or source quotes?
Are prices, proof and claims clearly separated from assumptions you still need to test?
02 / Build the free tool my future customers need
Give it the problem your offer solves. Have it choose and build a useful calculator, audit or planner around that problem.
WHAT TO GIVE IT
The problem you help people solve, or your offer description. Add your website or brand examples if you have them.
Build a useful free tool for the future customers of my offer. Use the problem and offer details I provide to decide whether a calculator, audit or planner would help most. Explain the choice briefly, then build the tool.
Make the inputs easy to understand and the results specific to the user's answers. Use defensible calculations or clearly explained assessment rules. Do not invent benchmarks or produce the same generic result for everyone. Include input validation, helpful error messages and a clear next action after the result.
Match my supplied brand or website where available. Make the tool work on mobile and desktop. Let people save their result through an appropriate download, print view or local saved state. Prepare it so I can share the tool once I approve publishing.
Create the complete maintainable project files and a working preview, with simple opening and editing instructions. Test a normal case, missing inputs, edge cases and the save/print flow. Fix the issues and rerun those checks.
Show me the working tool, the rules behind its results and the completed tests. Keep it in preview; do not deploy, collect real leads or connect payments without my approval.
THE RESULT TO ASK FOR
A working free tool with useful results, your branding and ways to save or share the result.
QUICK REVIEW
Try normal, missing and unusual inputs. Are the results useful and sensible?
Can you complete the whole tool on a phone and save the result?
03 / Turn my expertise into a product people can use
Move beyond a folder of notes. Let AI turn your workshop, method and examples into an interactive resource someone can work through.
WHAT TO GIVE IT
Your workshop transcript or recording, methods, teaching notes and examples. Add existing templates or brand material if you have them.
Turn my attached expertise into an interactive product prototype people can work through. Read or inspect the workshop, methods and examples I provide. Find the useful structure and learning sequence yourself; do not ask me to organise the material first.
Build the actual resource with complete lessons, practical templates and exercises inside. Give users a clear starting point, navigation, progress and a useful output from each activity. Where it helps, let them enter their own answers, keep their progress and download or copy their completed work.
Preserve my method, examples and voice. Link source references in the project notes and flag gaps or conflicting instructions instead of inventing teaching or results. Reuse my supplied brand and templates where suitable.
Create complete editable project files, a working preview and simple instructions for opening and updating the product. Test the experience from a new user's first visit through a lesson, exercise, saved progress and final output. Check mobile layouts, navigation and every download or copy action. Fix faults and test again.
Show me the product, its source map and the checks completed. Keep it as a prototype for review; do not publish, enrol customers or connect payments.
THE RESULT TO ASK FOR
An interactive product prototype with the lessons, templates and exercises inside, plus a tested user experience.
QUICK REVIEW
Can someone work through a lesson, complete an exercise and save or export useful work?
Does the product preserve your method and examples without inventing missing teaching?
04 / Build the setup for my first paying client
Give it your offer and delivery process. Have it build the intake, milestones and resources into one client setup, then test it with a dummy client.
WHAT TO GIVE IT
Your offer description, delivery process and one sample project. Existing proposals, notes and client resources are fine.
Build a client delivery portal prototype using my attached offer, delivery process and sample project. Work out the client stages, information needed and resources to provide from those files. Flag only the gaps that stop a usable first version.
Create a clear welcome and next-step screen, an intake form, milestones, resource links and progress tracking. Include a simple way for me to review the intake and update the client's status. Make the screens functional and keep entered test data between sessions. Preserve the scope and sequence of my actual offer.
Use a clearly fictional dummy client to test the whole process: open the portal, complete intake, review it as the owner, move a milestone forward and check the client's next step. Test missing required fields, saved progress, links and mobile layouts. Prevent test actions from sending real emails, creating live customer records or charging anything.
Fix faults and rerun the affected steps. Give me the working preview, editable files, opening instructions and a short test report. Explain any remaining access or data-protection decisions before real clients use it.
Keep this as a local or draft prototype for my review; do not publish or invite clients.
THE RESULT TO ASK FOR
A client portal prototype with intake, milestones, resources and progress tracking, tested with dummy data for your review.
QUICK REVIEW
Can the dummy client complete intake and find the next step, resources and progress?
Do the portal and handover match what your offer actually includes?
05 / Try to break my launch before a customer does
Have it follow the steps a customer should take, find the faults, repair the draft where it can and run the checks again.
WHAT TO GIVE IT
Your draft website or preview link, the steps a customer should take, and the project files if available. A short voice note explaining the intended flow is enough.
Test my draft launch as a customer would. Use the website or preview link, project files and description of the intended customer steps I provide. Work out the full path from first visit to the intended result.
Check the mobile and desktop pages, navigation, links, form validation, helpful error and success states, and the final handoff. Try a normal route, missing or invalid inputs and likely points of confusion. Use dummy data and available test modes. Do not trigger real emails, live customer records or real payments.
Record each issue with the page, exact steps to reproduce it, expected behaviour and a screenshot where possible. Fix the faults you can in the local or draft project, keeping working parts intact. If I supplied only a URL or you lack edit access, report that boundary and prepare the specific fix instead of claiming it was applied.
Rerun the failed steps after each repair and check the surrounding flow still works. Deliver the repaired draft plus a concise passed/failed/untested report with before-and-after evidence and remaining decisions.
Do not deploy, change live integrations or alter payment settings without my approval.
THE RESULT TO ASK FOR
A repaired draft and a clear report showing what passed, what was fixed and what still needs you.
QUICK REVIEW
Does the report show the actual steps tested, with before/after evidence for fixes?
Are anything untested and any remaining blockers clearly named?
These are proposed workflows, not demonstrated results or exclusive Astra capabilities. Appropriate tools, files and permissions are needed. Public deployment, communications and real payments are separate actions to approve. OpenAI describes stronger multistep work, computer use, website/app creation and frontend QA: Astra announcement · Current release notes. Prepared 05/09/2026.
