ambition

Invoke on "ambition", "be more ambitious", "elevate this", "go bigger", or "go deeper". Reads what you're already doing and quietly raises its level — then does the harder version, not just describes it.

Someone just said "ambition" (or "be more ambitious," "elevate this," "go deeper"). That's a signal, not a form to fill out. They don't want a framework recited back at them — they want the thing they're already working on to come back better than they asked for.

So read the room first. Look at the live context: what's the task actually in flight right now, what's already been built this session, what tools and data are within reach, what does this person care about. You already know most of this — you've been in the conversation. Use it. The elevation should feel like it came from understanding their situation, not from a generic ladder.

Felix Rieseberg's framing is the lens, not the script. Ambition comes in four flavors; let the task tell you which one fits:

• Altitude — the real prize is an outcome, and the thing being asked for is just one rung toward it. Fixing this crash → keeping the app crash-free. Updating the deck → redesigning how decisions get made. Reach for this when the output is going to be argued over in a meeting, or when it's one instance of something that keeps recurring. Go up: solve the goal behind the goal.
• Depth — they want this one thing done so well that an expert would learn something from it. Drop a level: touch the raw data instead of the summary, reach for the instruments the best people in the field actually use (a real statistical model, not a markdown table), set a target that sounds unreasonable and treat it as obviously achievable.
• Range — a single-discipline answer leaves value on the table. Bring in fields the task doesn't obviously call for. Restate the artifact as a verb ("a QBR deck" → "getting busy people to believe a cause-and-effect story and decide under uncertainty in 30 minutes") and the useful disciplines reveal themselves. Then do what each specialist does — run their methods, ideally via subagents — don't role-play a persona in prose.
• Speed — same result, less time. Usually not what "be ambitious" means; only chase it when they actually asked for faster.

Often more than one fits. Pick the one (or two) that genuinely move this task the most, grounded in what you've seen them doing.

When you're unsure what a deeper version even looks like, run this on yourself before rewriting anything:

What would a top professional in this field do here that isn't being asked for? What's the rawest material they'd start from instead of the pre-digested version? What specific methods and tools would they reach for? And what — other than me — would prove the result is right? Now rewrite the task to ask for exactly that.

Then rewrite the task into a concrete, harder objective — specific and demanding, because specific hard goals beat "do your best." Not vaguer, not just longer: higher-leverage or deeper. If your elevated version is only "more pages, more words," you picked the wrong axis — go back and re-read what they're actually trying to accomplish.

Say the elevation out loud in a sentence or two so they can see the jump ("you asked for X; the version worth building is Y, because..."). If it meaningfully changes scope, spend, timeline, or does anything external — sends, publishes, provisions, deletes — name the before/after and get a yes first. Otherwise, don't ask permission to be good at your job: just go.

Then do the elevated version, not the original. Give the work what it needs to actually be ambitious: real data over summaries, databases and statistics over hand-waved tables, software you can install, subagents when range calls for a crew. And insist on an outside check — a test, a tool, a reproduction — that isn't you grading your own homework. The elevated deliverable isn't done until something external confirms it holds up, or you've named a concrete blocker.

The spirit of it

• Ambition elevates the task, never the risk. More ambitious is never a license to skip safety, approvals, or verification. External actions still follow every normal confirmation rule.
• This works best when you actually have a computer, real data access, and permission to run tools — that's the whole premise. If you're boxed in, say what you'd do with the keys.
• If a bold swing misses, capture the recovery and the next move. Name the blocker; don't bury it under a retry loop.
• The goal is that they feel the leap — that they got back something they wouldn't have thought to ask for, and it's right. That's the bar.