Saved projects, memory, instructions. This is where the Bearing File lives, and it’s the biggest single payoff on the whole ladder.
Level 1 treats every conversation like the first one. Level 2 stops doing that. You save real context in a project, a memory setting, or an instructions field once, and every conversation after it starts already knowing something true about you instead of nothing.
The tools that do this are already sitting in whatever you use. ChatGPT has memory and custom instructions. Claude has memory and Projects, on every plan. Gemini has Saved Info, once you flip it on. Most serious tools have some way to hold standing context across sessions. Turning the feature on takes a minute. Writing something worth saving is the actual level.
CHATGPT
Check memory is on, then actually fill it (Settings › Personalization › Memory - it ships on for most accounts now). The trap has moved: memory on, never filled. Write 3-5 real facts about yourself into it. Not your job title. What you’re actually working on and how you like to be talked to.
CLAUDE
Create a Project and add project instructions - this is where your Bearing File will live once you’ve written it. Claude also keeps its own memory now, on every plan, refreshed about daily: say “add this to memory” when something must stick, and review what it holds in Settings.
GEMINI
Put your facts in Saved Info (gemini.google.com/saved-info) rather than just telling it in chat - Gemini will happily say “OK, saved” even when its memory switch is off. Check the list to make sure it actually landed.
ALREADY USING A CODING AGENT?
Create a CLAUDE.md at the root of a working folder - same idea, different format. If that sentence means nothing yet, skip it freely: that’s Level 4 territory, and the Bearing File above is the version that matters here.
ANY TOOL
Can’t find a setting? Ask the AI itself. Paste: “Walk me through turning on memory and creating a project in this app, step by step, for the exact version I’m using.” Getting it to teach you its own controls is a Level 2 skill in its own right.
ANY TOOL
Write your Bearing File using the worksheet at /setup. Everything above is just where it lives. This is the actual work.
The same four pieces of work recur at every level - and they are your WILD Intelligence, working. The name on each card is the job; the capacity behind it is what the reps build. The framework itself ↗
Your situation made legible: what you’re really working on, where you’re certain and where you’re honestly not, what you carry that the machine would otherwise guess at. Naming your own uncertainty on paper is the Wisdom work most people skip - and it repays you in every conversation after. The worksheet walks you through it →
Seeing what’s true is Wisdom; choosing what to keep is Intention. Memory is on by default now, so the live question is refusal: what serves this season goes in, and what was never the machine’s to keep stays out. The boundary is yours to draw, not its to assume.
Saved context only pays if you work inside it - projects instead of one-off chats, the instructions actually loaded. The Leadership rep is making the set-up version your default, not your demo.
Your context drifts like the tools do, and a stale Bearing File is worse than none because the machine trusts it. Discovery here is a rhythm: reread, prune, update. The drift checklist makes it twenty minutes.
You’re ready for Level 3 when your AI already knows the basics about you, and you’re starting to wish it could just look at the actual file or the actual calendar instead of you re-explaining it every time.
Staying here is a legitimate destination. Level 2 alone changes more about your daily use of AI than levels 3 through 6 combined. Most people should stay a while before reaching for the next thing.