Open Settings and sign in
Do this before your first real session so the app can load your current model catalog.
Guided Tutorial
Your first good session, start to finish.
This walkthrough is for normal users. It shows you what to click, what to expect, and what each part of the app is for without turning the setup into homework.
- Written for first-time users
- Covers Analysis, Reports, Capture, and Reviewer
- Short steps with clear expectations
- Built for the current v2 desktop flow
Before you start
Install the Windows app, launch it, and keep your browser available for GitHub sign-in.
What you will learn
How to move from Chat to Critic, open Analysis, generate Reports, attach evidence, import research, and search saved records.
Best outcome
You finish with one session, one critique pass, one analysis read, and one grounded report you can reopen later.
Contents
Jump straight to the part you need.
Start here
Quick start
The shortest path from install to first useful result.
Go to quick start Step 1Sign in and check models
Use Settings first so the app is connected before your first real turn.
Open sign-in guide Step 2Start in Chat, move to Critic
Clarify the idea, then pressure-test it without opening a new workflow too early.
See the lane flow Step 3Use Questions properly
Keep the gaps visible, answer them when ready, and avoid losing the hard parts.
Open question guide Step 4Open Analysis
Use Focus, Compare, and All views to see what kind of pressure the argument is under.
See analysis walkthrough Step 5Generate Reports
Turn the saved record into a session overview, contradictions report, or research summary.
See report steps Step 6Capture and attachments
Bring in screenshots, crops, and files when text alone is not enough.
Open capture guide Step 7Import outside research
Use Reviewer so external evidence is checked separately from ordinary chat.
Open reviewer guide Step 8Records and Settings
Find exact saved facts later and personalize the workspace without breaking the flow.
Open records guideQuick start
The shortest useful path through the app.
Use one running example all the way through the tutorial
If you do not want to invent your own case while learning, use this one throughout the tutorial:
Running example
I am considering leaving my product job in six months to start a small consulting practice focused on B2B SaaS onboarding. I want to know whether this is a good decision, what assumptions I am making, and what evidence I still need.
Why this works well
It is concrete enough for Critic, open-ended enough for Questions, and realistic enough for Analysis, Reports, Capture, and Reviewer imports.
Start one Chat session
Explain the idea in plain language and get it into shape before you attack it.
Send it to Critic
Use the same session so questions, contradictions, analysis, and reports stay grounded together.
Open Analysis
Use Focus for one lens at a time or Compare when you want one-vs-one pressure testing.
Generate one report
Use Session overview first. It is the fastest way to see whether the work is becoming coherent.
Come back later
Your session, questions, reports, attachments, and settings persist locally on your machine.
Step 1
Open Settings first and finish sign-in
Do this
- Open the app.
- Click the Settings button.
- Click Sign in with GitHub.
- Finish the browser approval step.
- Return to the app and wait for the catalog to load.
What you should see
Settings should show your selected model, the current account catalog, and a short explanation if some models are unavailable.
See a concrete example
A simple readiness check for a new user:
Ready means
I can see a selected model, I can open the model list, and Settings no longer looks like a sign-in screen.
- If you can choose a model, you are ready to start a real session.
- If Settings still looks empty or limited, resolve that first instead of guessing later.
Step 2
Start in Chat, then move the same session to Critic
Do this
- Stay in the Chat lane.
- Describe the idea, decision, or claim in plain language.
- Ask the app to sharpen, organize, or clarify it.
- When the wording feels honest enough to challenge, click To Critic.
- Ask Critic to test assumptions, contradictions, missing proof, or weak definitions.
Best practice
Do not open a fresh session just to move into Critic. Keeping one continuous session gives Analysis and Reports a better record to work from.
See a concrete example
Example Chat prompt
I am considering leaving my product job in six months to start a small consulting practice focused on B2B SaaS onboarding. Help me make this idea precise. What exactly is the claim, what assumptions am I making, and what would success look like in year one?
Example Critic prompt
Pressure-test this plan. Where am I relying on wishful thinking? What would have to be true for this consulting idea to fail? Which assumptions need evidence before I act?
Step 3
Use Questions as your working backlog
Do this
- Watch which questions appear after Critic turns.
- Answer them directly in place when you have enough evidence.
- Mark questions resolved when the gap is closed.
- Archive them when they are no longer worth working.
- Reopen older questions from history if the issue becomes active again.
See a concrete example
Using the consulting example, good active questions might look like this:
Example question list
- Do I already have proof that buyers will pay for this service?
- What specific problem am I solving better than an internal operations lead?
- How many months of savings do I need before leaving salaried work?
- Am I assuming demand because a few people expressed interest?
Step 4
Open Analysis when you need structure, not just more talk
Do this
- After a substantive Critic turn, open the Analysis panel.
- Use Focus when you want one lens at a time.
- Use Compare when you want the primary read plus Comparison snapshot and Comparison lens detail.
- Use All when you want a compact board view of the whole analysis.
- Use the Context Manager when you want built-in or custom contexts to shape the alignment read.
If Analysis looks empty
That usually means the argument is still too thin. Go back to Critic, send a fuller version, then reopen Analysis.
See a concrete example
Example Focus use
Open Focus and look for the strongest uncertainty first. In the consulting example, that might be whether demand is an empirical problem, not just a motivational one.
Example Compare use
Use Compare when you want to see how two lenses read the same idea differently, such as a practical lens asking "will anyone buy this?" versus a systems lens asking "what dependencies or bottlenecks make the plan fragile?"
Step 5
Generate reports from the saved record
Do this
- Open the Reports panel.
- Generate Session overview first.
- Use Contradictions when you want failure points highlighted directly.
- Use Research summary when outside material is central to the work.
- Reopen older saved runs from history if you want to compare how the session changed over time.
See a concrete example
Use this report order
1. Session overview: when you want the full state of the decision.
2. Contradictions: when you suspect your plan sounds confident but does not hold together.
3. Research summary: when outside evidence is now influencing the decision.
For the consulting example, the contradictions report is useful right before making a real yes-or-no decision.
Step 6
Use Capture and attachments when text is not enough
Do this
- Click Capture to take a cropshot from the desktop drawer.
- Drag files into the composer or click Attach files.
- Check the Ready to send strip before you submit the next turn.
- Ask for text extraction when readable text is the main point.
- Ask for direct image inspection when layout, charts, screenshots, or visuals matter more than OCR.
See a concrete example
Example capture prompt
I captured a screenshot of my revenue spreadsheet. Inspect whether this forecast depends on unrealistic client conversion assumptions.
Example attachment prompt
I attached a competitor service PDF. Pull out the claims that matter most for whether my consulting offer is actually differentiated.
Step 7
Import outside research into Reviewer
Do this
- Open Settings.
- Expand Manual fallback and import tools.
- Open Research import.
- Turn on Allow GPT-Researcher imports.
- Paste GPT-Researcher JSON or bullet output.
- Click Import research.
- Switch to the Reviewer lane and ask what the imported material actually proves, what it misses, and what still needs checking.
See a concrete example
Example imported research
- B2B SaaS onboarding budgets increased in mid-market software firms.
- Founders complain that internal onboarding teams often lack implementation discipline.
- Several case studies show retention gains after structured onboarding redesign.
Example Reviewer prompt
What in this imported research actually supports my consulting idea, what is only generic industry background, and what would still need direct proof from my own market?
Step 8
Use Records for recall and Settings for control
Use Records when you want exact recall
- Ask for saved questions, contradictions, reports, sessions, or captures.
- Use quick prompts when you want fast lookup over stored records.
- Turn on the interpretive layer if you want a grounded explanation on top of the exact result.
- Pick Studio, Slate, or Forest theme.
- Choose compact or comfortable spacing density.
- Turn auto-naming on or off.
- Review model access, token source, and available models.
- Enable or disable follow-up question generation.
See a concrete example
Example Records queries
List unanswered questions from this consulting session.
Show contradictions found in this session.
Show saved reports for this session.
Example Settings use
If the session is becoming intense, keep the model the same and lower the mental load by switching to a calmer theme or comfortable density. If the app is asking too many new questions, turn question generation off for a while.