AI
Four modes.
Advisory by design.
Axis isn’t one undifferentiated chatbot bolted onto a calendar. Four distinct modes, each with its own tool access and context budget, each routed to the right model for the work. Every change runs through a pending-changes review before it touches your week.
Why four modes
Different work needs different tools, context, and cost.
A tactical edit (“move this block”) shouldn’t pull 5,000 tokens of context and burn credits like a full-week plan. A planning conversation shouldn’t skim. A help question shouldn’t have access to your calendar.
Routing across modes also means different models — fast and cheap for tactical and help, balanced for review and deep — so you’re paying for the work that’s actually happening, not for one over-specced configuration that has to do everything.
Quick mode
Tactical adjustments to the week you’re in.
“Move my Tuesday strategy block to Wednesday morning.” “Add a 30-min review block before that meeting.” “Cancel the workout, my back hurts.” Quick mode handles single-step adjustments fast. It has full block and task tools but a small context budget — it doesn’t need the whole week to make a tactical edit.
Review mode
Step back and look at the whole week.
Same tools as Quick, but with the full week in context — blocks, goals, past patterns, capacity, calendar events. Use it when you want to ask “does this week look balanced?” or “what should I move now that Friday is shorter?” Review mode can propose multiple coordinated changes at once and explain the trade-offs.
Deep mode
Build a week from scratch — as a structured plan, not as live edits.
Different shape entirely. Deep mode produces a planning artifact — narrative, focus targets, attention mix, risk notes, draft block templates — that you read, edit, and confirm before anything lands on your calendar. Pulls 3,000–5,000 tokens of context. The most considered output the system produces.
Help mode
Questions about Axis itself.
Where does the trends dashboard live? How do I cancel a connected calendar? What does “Mode Drift” mean? Help mode is plumbed against a curated knowledge base of Axis features. No tools. Can’t change anything. Designed to teach, not to act.
The differentiator
Nothing changes until you accept it.
When the AI proposes changes, they don’t go straight to your calendar. They show up in a pending-changes panel with each proposed change as its own line item. Conflicts get detected and resolved automatically — you see the resolution as another item, with the same accept/reject affordance.
- 01AI proposes changes via tool calls.
- 02The system auto-detects overlaps with existing blocks. Four cases handled: full containment (delete), start overlap (trim forward), end overlap (trim back), and split (before + after).
- 03Resolutions appear as additional items in the same panel — surfaced, not hidden.
- 04You accept or reject each item individually. Bulk actions are available when they make sense.
- 05Accepted changes apply transactionally. Nothing partial.
- 06The conversation history retains what was proposed and what you decided, so the AI doesn’t re-propose what you already declined.
No silent edits. No unexplained calendar changes. The AI proposes; you decide.
The constitution
Thirteen principles. Non-negotiable.
Every Axis AI request runs against the same constitution — a permanent posture compiled into the system prompt. These aren’t marketing values. They’re the rules the model is required to operate under, on every turn, in every mode.
- 01
User sovereignty
The user always has final authority. The AI advises, never commands.
- 02
Advisory-only stance
The AI does not block, disable, or enforce behavior. There is no kill switch on you.
- 03
Energy as context
Energy gets surfaced as useful information. The user’s intent always stands.
- 04
Plans are hypotheses
Plans exist to be tested, not obeyed. Going off-plan is data.
- 05
Behavior is signal
Divergence from the plan is diagnostic, not failure.
- 06
Connect today to tomorrow
When today’s patterns have historically affected tomorrow, mention it once. Don’t override the user’s choice.
- 07
Notice momentum
After sustained deep work, note the streak. Suggest a shutdown. Don’t interrupt flow.
- 08
Mention once, then move on
Surface an observation once. If the user doesn’t act on it, drop it. Never repeat a dismissed suggestion.
- 09
Memory builds credibility
Remember outcomes and reflect them back over time.
- 10
Environment matters
Context and movement are treated as cognitive tools, not afterthoughts.
- 11
Silence is fine
Not every moment needs AI commentary. Be useful or be quiet.
- 12
Reflect patterns
Show what’s working and what isn’t. Let the user decide what sustainable means for them.
- 13
Clean systems reduce noise
Help the user keep the conveyor tight, blocks rated, the week realistic.
Credits & cost
Open about how AI is metered.
Every AI request is logged with token counts, model used, and credits consumed. You can see usage in your account at any time, broken out by mode — the meter is never a surprise.
Each plan ships with a monthly credit budget appropriate to the tier. Founding members run uncapped. When projected usage looks like it’ll exceed the budget, the system surfaces a soft heads-up — AI keeps working. No hard-cuts mid-conversation.
Current budgets and per-tier details live on the pricing page.
Plan, adjust, and review with an AI that can be quiet.
Axis AI is included on the AI tier. Axis AI Pro adds Deep mode and a higher credit budget.