Prompting

Three Workflows That Turn GPT-5.6 Sol Into a Whole Team

Three workflows that turn GPT-5.6 Sol from a chatbot into a whole team. The exact prompts for all three are below.

6min read
4prompts to copy
5sections

The idea in one line. GPT-5.6 Sol is OpenAI's new flagship. Most people will open it, type a question, and use the strongest model they have ever shipped like a search box. The people getting a real edge run it as a system: they let it spawn a swarm, they route the cheap work to a cheaper model, and they let it drive their tools while they are gone. Here are those three workflows, with the prompts you can paste today.

Workflow 1Ultra mode: run a swarm, not one chat

What it is.Ultra mode is a multi-agent system built into Sol. Instead of grinding through your task in one long reasoning chain, Sol spins up several subagents that each take a different piece, work in parallel, and merge into one result. It is what takes a job that would eat a person a week and hands it back in minutes.

When to use it.Reserve it for big jobs that naturally break into parts: a full competitor teardown, a research report, refactoring a codebase, producing a whole set of assets. For quick questions, skip it, it is slower to spin up and it costs more. Ultra mode shines when the work is large and splittable.

The move that makes swarms good. Do not just say 'use ultra mode.' Give it a job that clearly decomposes, tell it to split into parallel subagents, say what each should own, and tell it to merge and flag conflicts. Structure in, quality out.

PASTE THIS · THE ULTRA-MODE BRIEF
Use ultra mode. This is a big job, so split it across parallel
subagents instead of doing it in one pass.
Goal: [the outcome you want, e.g. a full teardown of [company]:
positioning, pricing, product gaps, and 3 ways to beat them].
1. Break this into independent sub-tasks.
2. Assign one subagent to each and run them in parallel.
3. Merge everything into a single [report / doc / plan].
4. Flag any place the subagents disagree so I can decide.
Use max reasoning effort. Take the time to get it right.

Turn it on. In ChatGPT, pick GPT-5.6 Sol and switch on Ultra mode before you send the prompt. In the API or Codex, request Sol with ultra mode plus max reasoning effort. If the toggle name or placement is different at launch, look for the 'ultra' or 'subagents' option on the Sol model, it is a first-class feature, not a hidden setting.

Keep reading

Free, for an email.

Unlocks 4 more sections, 3 copy-paste prompts, the document version, and every other guide on the site, for free. Enter your email once to keep reading.

19,000+ follow where these guides come from.No spam. Unsubscribe in one click.

Read next