Use ChatGPT for founder work from the first question: explore directions, test assumptions and decide what deserves your next investment. If the evidence supports continuing, connect what you learn to MVP scope, marketing and first-user experiments.
This guide is for a solo software founder, whether you are exploring ideas, testing a hypothesis or working on an existing product. The examples are hypothetical teaching examples, not customer case studies or measured results.
Start with the context, not the perfect prompt
A persuasive plan can still be wrong for your company. A launch plan that assumes an existing audience is unhelpful if you have none. An MVP that requires three specialists is unhelpful if you build alone after work.
Before each task, give the assistant a compact record of the product, intended customer, founder resources, evidence, and current decision. Label the difference between a fact and a hypothesis. “Agents need this” is a hypothesis; “three interviewed agents described the same workaround” is evidence with a sample limit.
Your context should answer four questions: Who is this for? What progress are we trying to create? What do we actually know? What can this founder do next? Build that context file to capture what you know and what you still need to learn.
Connect the founder workflow through decision gates
Already have a working product? The existing-product intake guide explains how to collect sourced context from its website, repository, docs and current offer. This workflow is available to every User and uses prepaid MCP credit.
| Stage | Useful AI work | The human evidence gate |
|---|---|---|
| Explore | Compare directions from your experience, constraints and customer access. | Which problem deserves a closer look? |
| Shape | Clarify a problem and customer hypothesis. | Can you reach people who experience this situation? |
| Validation | Organize alternatives, risks, and experiments. | Did customers show the problem and take a meaningful next step? |
| MVP | Scope one outcome and its acceptance criteria. | Will the smallest version test the remaining uncertainty? |
| Build | Turn the agreed scope into a reviewable implementation. | Does it work for the intended user, including failure paths? |
| Customers | Draft a focused channel experiment and outreach. | Do people activate, return, and consider paying? |
Each gate changes the next brief. If interviews reveal that the painful task is chasing unanswered inquiries, the MVP should solve that task. It should not preserve a generic CRM feature list because an earlier chat wrote one.
A gate can support a bounded test, a changed hypothesis or a no-go. A research report alone does not justify a full build. Save the reviewed evidence, the decision and what would make you reconsider. Keep rejected options and brainstorming in working notes, rather than promoting them into current product facts.
Connect a product decision to its marketing consequences
Suppose customer evidence supports a reminder feature for freelance designers who lose track of proposals. After accepting its scope, consider how that problem should appear on the landing page, in a short social post, in an article about proposal follow-up and in the criteria for finding relevant pilot users. These are related tasks to choose, rather than a requirement to publish every format.
In 1PC, Product Owner checks saved state before selecting a workflow. It can suggest relevant marketing, social, SEO or lead-research work when the product decision affects those areas. A suggestion does not authorize sending outreach or publishing. If the current feature scope is still uncertain, record that gap before making launch claims.
The SaaS feature-launch marketing plan turns that connection into a practical brief: one customer problem, consistent assets and a bounded channel experiment.
Keep the operating loop useful after the first launch
An existing product needs a recurring way to choose the next action. Use the weekly product review worksheet to compare saved evidence, unfinished work and available capacity. A draft, an accepted decision and a verified customer outcome should not receive the same status.
If you repeatedly explain your own review checklist, turn it into a private custom product skill. Publish the method on use1pc, use it with current product context and save the result for the next task. These three guides cover different parts of the same loop: decide, execute and improve your method.
Give your users product-aware support
Customers also need current context. A Product owner can enable Customer support and publish a separate public Markdown folder. Guests use the 1PC Support connection without a 1PC account. The support documentation guide covers publication and private-knowledge boundaries.
A worked example: AgentCRM
Suppose you know independent real-estate agents, have 20 hours a week and a $500 monthly budget, but no settled idea. One direction to investigate is inquiry follow-up. Treat lost inquiries as a hypothesis rather than deciding to build a CRM.
- Idea Intake: describe the problem, intended customer, current alternatives, founder advantages, and unknowns. Save the brief as a scoped Strategy reference.
- Validation: test the problem before building. Ask agents about their last missed follow-up, then offer a manual pilot. Save the evidence and next experiment as scoped Research references.
- Decision: go for a bounded test, change the hypothesis or stop this investment. Missing evidence may justify interviews without justifying an MVP.
- Planning: only if the evidence supports a software experiment, scope a simple inquiry list and next-action reminder. Put unrelated CRM features outside the first release.
- Customer learning: observe whether the pilot fits the agent's day. Update your product overview with what changed.
The transition is the useful part: a new task reads the latest customer hypothesis and constraints, rather than inferring them from an old transcript. Use the validation guide, then the MVP planning guide, and finally the first-customer playbook.
Use ChatGPT Projects or a product context layer?
ChatGPT Projects can share uploaded files, instructions, and sources across related chats. They are a useful starting point when you maintain the working context yourself. See OpenAI's Projects documentation.
1PC keeps founder context, versioned Knowledge, decisions and unfinished Tasks. Every User can use the available strategy, research, design, engineering, marketing and customer skills. Publish private methods in your workspace and load them through Product Owner. Product membership does not share a private skill library. Accepted private MCP calls consume prepaid credit. Connect and run your first workflow.
Choose the simplest system you will keep current. If a manually maintained project meets your needs, use it. Consider a structured context layer when the same founder facts and product decisions need to be read and updated consistently across repeated tasks.
Know the limits of the system
If your next task uses another client, explore the Claude founder workflow or the Kimi founder workflow. The MCP overview explains other connection paths. Continue from saved product records, and use each client's setup instructions rather than assuming the ChatGPT plugin is installed there.
Persistent context does not make assumptions true, guarantee customer demand, or remove the need to review output. A saved document can be outdated. An assistant can overlook a constraint. Require it to summarize the context it used and identify conflicts before making a recommendation.
The 1PC backend stores your profile and current saved files, rather than automatically storing raw conversations. Knowledge files retain revision history, alongside decision history and structured progress. A deployed, compatible connection is needed to use these files from ChatGPT; account creation alone does not install a plugin.
Start with one product and one repeatable loop. Without a clear idea, begin with your experience, resources and the people you can reach. Compare candidate directions, then clarify and save a brief for validation. For a working product, collect its current context, choose a concrete improvement, save the result and continue in a fresh conversation. The next task should start from the actual saved records.
Choose one founder question to work on.
Start with 100K free tokens. Every User can connect saved evidence to MVP, marketing, sales and onboarding work using prepaid MCP credit.
Create your free workspace ↗