// release notes

Everything we’ve shipped.
Every decision we’ve made.

15 releases · 8 weeks · 1 goal: make 5 minutes a day matter.

May 12, 2026
latestv0.15

Multi-Everything

Your role isn't single-threaded. Your sessions shouldn't be either.

  • FeatureMulti-industry — pick up to three industries in onboarding (was one). Every selection shapes your lesson hook, news of the day, and exercise scenarios. “Founder building an EdTech marketplace with consumer plays” now teaches you for that exact intersection, not just one slice of it
  • FeatureFast-path that actually fills in — paste your one-liner bio in Q2 and Q3-Q8 reliably pre-select your role, industry, tools, and seniority. Tap Continue six times to confirm, or change what’s wrong. The previous version sometimes dropped extracted values when you clicked too fast
  • FixYour free trial now starts when you start — the 30-day clock kicks in the moment you finish onboarding, not the moment you sign up. If life pulls you away mid-flow for a week, you don’t lose a week of trial
  • FixExercise coach gives real feedback again — every submission now comes back with a specific “Did well / Missing / Try this” rather than a generic “response saved” placeholder. A thinking-model token ceiling was silently truncating the AI’s feedback; that ceiling is now set high enough for the full reply
  • FixThe Feedback button no longer hides Continue — onboarding’s bottom-right CTA is fully visible on every step. The widget reappears once you reach the dashboard, where you actually have something to feed back about
  • FeatureSettings reads like a profile, not a database — role, industry, seniority, AI fluency all show their human labels (e.g. “Tech / SaaS · Education”) instead of raw slugs. Mirrors how the admin and Learning DNA pages already render
May 7, 2026
v0.14

Fresher, Stickier

If your lesson on “frontier models” namedrops GPT-4o, the lesson is already stale.

  • FeatureEvery lesson now ends with Tomorrow’s Preview — a one-line teaser of what’s coming next. Closing the app feels like a chapter break, not a stop. (The same open-loop effect that makes you queue up just one more episode — but with concepts you actually want back tomorrow)
  • FeatureToday’s session references today’s AI — the knowledge-base backbone now refreshes against this week’s frontier-model releases, pricing shifts, and tool launches. Your lesson on “multi-model orchestration” mentions Gemini 3 Flash and Claude 4.7 Opus, not last year’s GPT-4o
  • Feature“Watch related” videos are actually related — the companion YouTube clip next to each lesson is now quality-gated for relevance, length (2–30 minutes), audience (10K+ views), and English audio. No more Precision Screwdrivers clip showing up next to a lesson on RAG
  • FeatureDaily session cap landed — 3 sessions per day, so the 60-session free-trial ceiling can’t be burned in a single weekend. Pace is part of the product
  • FeatureAdmin-triggered “Refresh now” — when a brand-new model drops mid-week (looking at you, GPT-5.5 Instant), an admin can re-prime the knowledge-base cache immediately instead of waiting for the weekly cron
  • FixSeniority badge no longer lies — selecting “Lead / Exec” in onboarding now shows “Lead / Exec” everywhere (profile, admin, header). A stale fallback was silently labelling every Lead user as “Senior”
  • InfrastructurePrivileged columns (trial state, wallet balance, admin flags) are now write-locked at the database layer — no client-side code path can modify them, even by mistake. The legitimate writes flow through auth-gated server routes
Apr 28, 2026
v0.13

Profile DNA, Deeper

Onboarding that listens, not interrogates.

  • FeatureOnboarding is now a guided conversation, not a form — Mochi walks you through 8 questions with a one-line “why I’m asking” before each, so you understand what every answer changes about your daily sessions
  • FeatureCinematic intro — Mochi narrates the arc from today to a few weeks from now, evolving as he walks the timeline, so the version of you you’re working toward is clear before you answer anything
  • FeatureBio fast-path — paste a LinkedIn-style summary or job description and the next several questions come pre-filled. Saves you six taps when it works, falls back to the regular chip flow when it doesn’t
  • FeatureThree new signals captured during onboarding — your bio in your own words, your industry’s specific slice (capital markets vs. fintech, biotech vs. provider), and your role’s flavour (sell-side vs. buy-side, etc.) — every one of them feeds your daily session
  • Feature“Other (type)” everywhere — your role, industry, sub-industry, tools, and goals all have a free-text option now. Whatever you do for a living is welcome, even if our menu doesn’t list it
  • FeatureThe Chat with Mochi now knows where you are — open it on the lesson, exercise, news, or quiz and it tailors its response to that context. The same question gets a different (better) answer depending on what you’re looking at
  • FeatureAsk the mentor for a quiz answer? It refuses, then turns the question back around with a hint that teaches you the concept. Your daily sessions stay yours to earn
Apr 21, 2026
v0.12

Trust Pass

If you can't trust the surface, the depth doesn't matter.

  • FixSignup now works reliably across every domain — a quiet captcha-layer issue had been silently blocking real users from completing signup since our custom-domain rollout. Fully resolved.
  • FeatureMochi greets you by name on the dashboard now, in the version of him that’s evolved alongside your progress — not a static placeholder
  • FeatureThe session view (lesson, exercise, news, quiz) now fills your screen properly on desktop instead of leaving large empty side gutters. Calmer to read, easier to focus.
  • InfrastructureYour onboarding answers are now strictly scoped to your account — sharing a browser, switching devices mid-flow, or signing up after another user signed out will never pre-fill someone else’s half-finished form into yours
  • FixIf you ever land in your dashboard with a missing profile (rare, but possible after auth changes), you’re now redirected back to onboarding to start clean — instead of being shown someone else’s Learning DNA
  • PerformanceWatch your daily session generate live — the loading screen now shows real pipeline stages with timing, plus a Mochi-at-laptop animation, so you can see your AI is actually working
  • FixFixed: rare scoring drift if you completed a session and hit retry simultaneously. Your AI Readiness Score is now atomic — one session, one update, always.
Apr 20, 2026
v0.11

News, Lessons, and the Managed Tier

Three of the most-asked-for features land on the same day.

  • FeatureManaged Tier — your first 30 days (or 60 sessions, whichever comes first) are on the house, no API key, no credit card. After the trial, bring your own key — wallet top-ups land in a future release.
  • FeatureWallet — refund-safe accounting on every credit and deduction is wired (top-up UI ships in a follow-up release). Per-session AI cost is shown in real time with no platform markup.
  • FeatureDaily AI News for your role — a new public newsroom drops a role-tagged article every morning, automatically deduplicated against what you’ve already read
  • FeatureLessons now render in semantic blocks — concept, example, exercise, recap, sources — instead of one wall of text. Way easier to scan on desktop, way easier to read on mobile.
  • FeatureFive new exercise kinds rolling out behind a flag — beyond open-text, you’ll start seeing prompt-craft, scenario defense, debug-the-output, and others matched to what the lesson is teaching
  • PerformanceReal-time progress during AI generation — the wait now shows you what’s happening at each step (“reading your profile…” → “building your Learning DNA…” → “saving…”) so 60 seconds feels like progress, not anxiety
  • FeatureYour lessons stay current automatically — the AI landscape backbone refreshes every week with the latest model releases, pricing shifts, and tool launches, so today’s session reflects this week’s reality
Apr 15, 2026
v0.10

Polish Pass

The little things that make the big things feel finished.

  • PerformanceBetter readability across the dark theme — every secondary line of text now has enough contrast against its surface to pass accessibility standards. No more squinting at muted info.
  • PerformanceOld feedback drafts now clean themselves up after 30 days — you won’t see a month-old half-typed thought the next time you click feedback
  • FeatureReset Account properly wipes everything now — onboarding answers, skill file, sessions, scores — and walks you back through a clean slate. No manual cleanup, no orphan data.
  • PerformanceLong admin actions (search, filter, paginate) stay snappy even with hundreds of users on the platform — under-the-hood query rewrites you’ll only notice as “it just feels faster”
  • FeatureOnboarding now has a clearer cancel + “taking longer than usual” signal if your AI provider is slow. You’ll never sit in front of a blank screen wondering whether something is working
  • FixFixed: a handful of small UI quirks (CSV export ordering on the admin user list, mobile column overflow on the dashboard, color drift on the mastery score badge after a recent retry)
  • FixFixed: the admin “Live Errors” panel now actually shows live errors instead of a “not configured” warning
Apr 14, 2026
v0.9

Beta Tooling & Resilience

Shipping to real users means handling real failures.

  • PerformanceSession loading now scales to hundreds of completed days without bloat — your dashboard timeline shows your last 30 days, and asking the mentor about anything older pulls that day's full content on demand
  • FeatureFirst-run feature tour — new users are walked through every part of the dashboard on their first login, so nothing feels unfamiliar
  • FeaturePersistent feedback button — available on every page so you can flag anything the moment you notice it, not after you've forgotten it
  • FeatureReset Account — beta users can wipe their data and redo onboarding cleanly, without needing to create a new account
  • InfrastructureSession generation is significantly more reliable — a edge case that could leave the loading screen stuck permanently after a failed attempt has been resolved
  • InfrastructureAI generation now has a hard 2-minute ceiling — if your provider is slow or unresponsive, the app recovers gracefully instead of hanging indefinitely
  • InfrastructureGenerating a session now shows a cancel button after 15 seconds and a clear 'taking longer than usual' signal after 30 — you're never left guessing whether something is working
  • FixFixed: completing onboarding could in rare cases redirect you to an empty dashboard with no Learning DNA. It now catches this, rolls back, and asks you to try again
Apr 10, 2026
v0.8

Production Hardening

The difference between a prototype and a product is what happens when things go wrong.

  • InfrastructureYour AI Readiness Score now updates correctly every time — a sequencing issue that caused score drift when retrying a completed session has been resolved
  • InfrastructureRapid clicks and page remounts no longer trigger duplicate AI generations — eliminates the edge case where two API calls could fire simultaneously and double-charge your key
  • FeatureCost transparency in mentor chat — you now see exactly what each conversation cost on your API key, not an estimate
  • FeatureGemini users: AI costs were previously being undercounted because reasoning tokens weren't included. This is now fixed — what you see is what you actually spent
  • InfrastructureYour Learning DNA is now protected against data conflicts — if you complete sessions from multiple devices or tabs, the most recent update always wins cleanly
  • FeatureErrors across the dashboard now tell you what went wrong and what to do — invalid API key, rate limit hit, network failure — each surfaces a clear, specific message instead of a blank screen
  • FeatureOnboarding form answers now save automatically as you type — refreshing the page or losing your connection no longer means starting over
  • FixFixed: if AI generation timed out during onboarding, the app previously sent you to the dashboard anyway with no Learning DNA. It now returns you to the API key step with your answers intact
Apr 2, 2026
v0.7

Public Architecture

Open source requires explaining your decisions, not just your code.

  • FeatureArchitecture page — full technical breakdown of how Day01.AI works: AI routing, skill file lifecycle, scoring model, BYOK security model
  • FeatureScroll-driven technical walkthrough animation (5 scenes, Framer Motion)
  • FeatureOpen source on GitHub under MIT License
  • FeatureInvite-only beta launch with closed-access gating via environment flag
  • PerformanceLanding page rebuilt: mobile-responsive nav, reduced bundle, semantic structure
Mar 28, 2026
v0.6

AI Readiness Score

Learning without measurement is just reading. We added measurement.

  • FeatureAI Readiness Score — composite metric across quiz accuracy, exercise quality, streak consistency, and phase progression
  • FeatureXP system: +15 XP for perfect quiz, +15 XP for substantive exercise, +5 XP for 7-day streak
  • FeaturePhase multipliers — later phases weight contributions higher (Hands-On: 1.5×)
  • FeatureStreak tracking with decay: 3+ inactive days reduce score by 1/day
  • InfrastructureScoring pipeline runs atomically — score update and session completion in a single guarded transaction
Mar 22, 2026
v0.5

Onboarding & Multi-Provider

Cold-starting personalization at sign-up, not after 5 sessions.

  • FeatureFull 8-step onboarding: role, industry, goals, learning style, experience level, timeline, self-assessment, API key setup
  • FeatureMulti-provider AI: Anthropic Claude, OpenAI GPT, Google Gemini — user selects and brings their own key
  • FeatureLearning DNA generated at onboarding completion — users arrive at their first session already profiled
  • FeaturePer-provider model fallback chain — if primary model fails, silently retries on next available model
  • FeatureAPI Keys Guide: step-by-step instructions with cost transparency (~$0.02/session, no subscription)
  • InfrastructureAES-256 client-side encryption for API keys — keys never reach the server
Mar 15, 2026
v0.4

Chat with Mochi

A lesson teaches. A mentor remembers.

  • FeatureChat with Mochi — persistent chat assistant with full session history context
  • FeatureMentor system prompt loads skill file + last 10 sessions on every turn
  • FeatureContext badge shows 'N sessions loaded' — transparency into what the mentor knows
  • FeatureMessage history stored per-user with 50-message cap and 4,000-char per-message limit
  • InfrastructureStreaming responses via SSE for all three providers (Anthropic, OpenAI, Gemini)
  • InfrastructureFallback to non-streaming with word-delay simulation for providers where streaming is unavailable
Mar 10, 2026
v0.3

Learning DNA

The core insight: AI tutoring shouldn't start from zero every session.

  • FeatureLearning DNA — a versioned markdown document capturing role, goals, curriculum, mastery signals, and gaps
  • FeatureDNA auto-updates after every completed session: topics mastered, weaknesses surfaced, next session pre-planned
  • Feature4-phase curriculum structure embedded in DNA: Foundations → Role-Specific → Hands-On → Stay Current
  • InfrastructureSkill file stored in Supabase with version column — full history preserved, conflicts resolvable
  • InfrastructureCurriculum generation uses user profile + industry context + previous session data as AI prompt inputs
Mar 5, 2026
v0.2

Structured Sessions

A lesson alone isn't a session. A session tests, applies, and closes.

  • FeatureQuiz module: 4-option multiple choice, instant feedback, explanation on correct answer
  • FeatureExercise module: open-ended written response with minimum length validation
  • FeatureMyth Buster: one AI misconception debunked per session, relevant to the lesson topic
  • FeatureSession state machine: lesson → quiz → exercise → complete
  • InfrastructureSession data stored in Supabase with completed flag, quiz_responses, and exercise_response columns
Feb 28, 2026
v0.1

Initial Release

Started as a tool built for one person who felt behind on AI.

  • FeatureAI-powered lesson generator: role + topic → structured 5-minute lesson
  • FeatureLesson format: concept explanation, real-world application, key takeaway
  • InfrastructureNext.js 15 App Router, Supabase Auth, Tailwind CSS v4, multi-provider AI abstraction layer
  • InfrastructureDirect browser-to-provider AI calls — no proxy, no usage data, no server-side AI handling
Feb 28, 2026 — day one