AI NATIVE ACADEMY Sandbox · State Frames

Four states. One conversation.

These are the load-bearing moments of the chat-first sandbox after the user submits their attempt: streaming (mid-evaluation), hint expanded (after pressing Hint 1), success passed (perfect attempt, lesson complete), and error / network failure. Stacked here for review; each is the same chat shell with different content rendered into the conversation column. Style B (Luminous Dark Glass).

№ 01 · STATE Streaming — mid-evaluation
Right after the user submits, before the rubric resolves
just now YOU 
You are a senior marketing analyst at a B2B SaaS company.

Given: our last 3 product launches under-performed against pipeline targets by 30–40%. We're preparing for Q3 and need to avoid a 4th miss…
AINA evaluating · 2.3s elapsed
Parsed structure
·
Scored elements
·
● Composing feedback…
· Hint check

Solid structure. You hit four of the five elements cleanly. Here's the breakdown|

Rubric · loading
7|/100
Calculating delta vs. attempt #1…
Clarity 18/20 ✓
Context scoring…
Output Format queued

● AINA is composing feedback. You can keep typing — your next attempt will queue.

№ 02 · STATE Hint expanded — nudge unlocked
User clicked "Hint 1 · Nudge" — hint inserted as a structured AINA message
Earlier in this exchange
You scored 72/100 on attempt #2 — strong on Clarity (18/20), thin on Context and Output Format. You asked for a hint.
AINA hint unlocked · level 1 of 3
1 Hint · Nudge Costs 0 points

Look at what's missing, not what's there.

You named the role and the task — the model knows who it's playing and what to do. But re-read your prompt and ask: would a brand-new analyst at this company have enough to start? What would they want to know first?

Try this

"Add 1–2 sentences of business situation before the task. Market, competition, timing — pick what matters most."

Hint 1 used · 2 remaining

Take a swing. I'll be here.

You are a senior marketing analyst at a B2B SaaS company.
Our category is sales-engagement tooling. We've launched two new pricing tiers in 18 months and a competitor (Outreach) shipped a similar feature 60 days ago. ←|
№ 03 · STATE Passed — lesson complete
After the user nails it: 95+ score, lesson marked complete, next-step CTA
2 minutes ago YOU · attempt #3 
You are a senior marketing analyst at a B2B SaaS company in the sales-engagement category…

Context: We've launched 3 products in 18 months. Outreach shipped a similar feature 60 days ago. Pipeline targets are owned by RevOps and updated quarterly.

Task: Identify the top 3 root causes our last 3 launches missed pipeline targets by 30–40%…

Success criteria: A senior PM should be able to act on findings within one week. No vague recommendations.

Return as: 1-page exec brief.
TLDR (3 sentences) → Findings (3 bullets each) → 1-week action plan.
AINA ▲ +24 from #1 · personal best

There it is.

All five elements present, all five strong. The Context and Output Format upgrades were the unlock — the model now has both the situation AND a definition of "good" to self-check against.

▲ Lesson 04 complete
96/100
Passed.
Trajectory
#1 → #2 → #2.5 → #3
✓
Role
19/20
✓
Context
18/20
✓
Task
20/20
✓
Constraints
19/20
✓
Output
20/20

Your AI Readiness Score moved from 68 to 71. Path locked at 44% complete.

Now — the part that matters

Take this skill to a real Tuesday task.

You came in hoping AI would help with your launch playbook analysis. The prompt you just wrote is exactly the one you'd use in production. Want to run it now? I'll save the conversation to your private workspace.

Next lesson · 12 min
Identity & Role
Define how AI should understand its persona inside your workflow.
№ 04 · STATE Error — evaluator unreachable
Network drop, rate-limit, or evaluator timeout — recoverable, your work is safe
25 seconds ago YOU 
You are a senior marketing analyst…
Saved · attempt #3 preserved
AINA · evaluator unreachable

I couldn't reach the evaluator just now.

Your attempt is saved — I have the full prompt locally and your draft in the editor isn't lost. This usually clears in 10–30 seconds. I'll auto-retry, or you can press retry yourself.

evaluator/score · timeout after 8.0s · retry 1 of 3 in 5s
Auto-retry in 3s…

If this keeps happening: something is wrong on our side, not yours. Status page · Try again later

— end of state frames —

All four states share the same chat shell, the same composer, the same sidebars. The differences are entirely what gets rendered into the conversation column. That's the whole point of the chat-first move.