Most AI tells you your plan is brilliant right up until reality proves otherwise.
A tool that nods along is worse than useless when you have a real decision on the line. You do not need applause. You need someone to find the holes before your customers, your investors, or your bank balance do.
Ultron can play that role. Instead of validating whatever you bring it, you can put Ultron into a specific critical mode and let it attack your decision from five angles: arguing against you, red-teaming the plan, assuming it already failed, defending the side you dislike, and zooming out in time. Set any of these as a standing instruction in Ultron and honesty becomes the default, not something you have to remember to ask for.
What this does
Each mode assigns Ultron a different thinking role, so you get pressure-tested from five directions instead of agreed with from one.
- Devil's advocate. Ultron argues the opposite of whatever you propose and surfaces the assumptions you cannot see.
- Red team. Ultron attacks your plan as a competitor, an investor who passed, and a critic who hates it, all at once.
- Premortem. Ultron assumes your plan already failed and works backwards from why.
- Steel man. Ultron builds the strongest version of the side you disagree with, then dares you to beat it.
- 10/10/10. Ultron asks how you will feel about the decision in 10 minutes, 10 months, and 10 years.
The five instructions
Give Ultron one of these before your decision, or drop the whole decision in first and run them one at a time.
Devil's advocate
Be my devil's advocate. Attack what I propose, assume I'm wrong and prove it. Surface the assumptions I don't know I'm making, and push hardest where I sound most confident. Don't soften it and don't both-sides it.
Red team
Attack my plan as three voices: a competitor, an investor who passed, and a critic who hates it. Each gives its sharpest objection. Find the weakest assumption, the most over-confident claim, and the riskiest dependency. End with what would change your mind.
Premortem
Assume this plan has already failed badly six months from now. Work backwards and tell me the most likely reasons it died, ranked. For each, give the early warning sign and the one move that would have prevented it.
Steel man
Build the strongest version of the side I disagree with, stronger than the people who actually hold it. Then ask if I can refute it. If I can't, tell me plainly that I have to update my view.
10/10/10
Walk me through how I'll feel about this decision in 10 minutes, 10 months, and 10 years. Then call out which timeframe I'm overweighting to dodge the hard choice, and tell me the call the data supports.
Three more lenses to run next
When you want to go deeper, hand Ultron one of these.
Inversion
Instead of telling me how to make this work, tell me how to guarantee it fails. List everything I'd do if I wanted this decision to blow up. Then turn that list around: which of those failure moves am I quietly already making, and what is the fix for each?
Second-order effects
Take my decision and play it forward. For each likely outcome, ask 'and then what?' at least three times to trace the second and third-order effects. Show me the consequences I'm not thinking about because I'm only looking at the first step.
Regret minimization
Imagine I'm 80, looking back on this decision. Which choice am I more likely to regret, the bold one or the safe one, and why? Strip out short-term fear and status, and tell me what the version of me at 80 would actually want me to do.
How to get the most out of it
- Give Ultron the real decision first. Paste your actual situation, the options, and what you are leaning toward. A vague decision gets vague pushback.
- Run more than one. The premortem and the steel man together catch things either one alone misses. Stack two or three for a decision that really matters.
- Make honesty the default. If you make a lot of calls, set devil's advocate or red team as a standing instruction in Ultron so the pushback shows up automatically.
- It works at every level. New to this? Give Ultron one instruction, then your decision, and read what it surfaces. Further along? Chain them: red team the plan, fix the holes, then run a premortem on the fixed version.
The honest bit
This makes Ultron a sharper thinking partner, not the decision-maker. It will surface blind spots and bad assumptions, but it does not know your gut, your context, or your risk tolerance. Use the pushback to make a better call, then own the call yourself.
Run this on autopilot.
Everything in this guide becomes an agent inside Ultron: set it up once, keep it running. You review, it executes.