FlowTimr
Earlier: Andris Timer, AI Voice Time Logger, and the voice-driven timer concept.
Track work, preserve what was accomplished, and organize time by client and project.
PROJECT LEDGER / FLOWTIMR
The work
remembered.
This is the evolving history of the Flow apps: where their ideas began, how their names changed, what actually shipped, and what each build taught me. No fake startup mythology. A reconstructed concept is labeled differently from a verified release.
PRODUCT LINEAGE / EARLIER IDENTITIES
The names changed.
The problems stayed.
FlowPic
Earlier: Media Brain.
Find and organize photos by content, metadata, location, and intended use without moving the originals.
FlowClp
Earlier: VAULT.
Organize, review, rediscover, and reuse video footage while editing workflows grow beside the library.
Source note: this lineage was supplied by Andris on 2026.09.29 from a ChatGPT reconstruction based on earlier conversations, a Notion project book, and saved development artifacts. It is historical source material—not proof that every discussed feature shipped.
JULY 29 PORTFOLIO SNAPSHOT / NOT ORIGIN DATES
The experiments were already becoming a fucking ecosystem.
The project book captured a much wider portfolio than the three core Flow apps. July 29 is the date of this snapshot—not a claim that every idea began that day.
FlowBlck — voice-driven day planning.
FlowSnip — capture something together with why it matters.
FlowStorm — persistent voice, vision, and project memory.
Flow Beat Sync Editor — automatic editing around music.
Client Portal / Agency OS
Smart Sites / Agent 3
AI Content Engine
Set To Roam App
Glacier / NationalPark.Guide
Smoke + AQI Forecast
Hiking Journal / Trail Story
Second Brain / Life Auditor
Endless Theory
SWRWRD
Local Directory Platform
OpenClaw
Andris.AI
SetToRoam.com
Separate histories: the voice-driven website and marketing builder discussed September 10, graffiti-game work discussed September 23, and customizable designer browser discussed September 26 sit outside the July project-book snapshot.
The timer was already becoming a memory system.
The earliest recovered concept asked for the client and project whenever a timer started, accepted voice, typed, or recent-project input, offered optional 15-, 30-, or 60-minute reminders, and ended by asking: What did you complete? Tasks could belong to work or personal life, with reminders and recurring routines surrounding the timer.
What it taught me: the useful product was never the stopwatch. It was the connection between elapsed time, work context, and a human-readable record of what actually happened.
Reconstructed original concept / precursor, not a confirmed codebase start dateAndris Timer became FlowTimr.
The naming moved through FlowTime and shortened variations before FlowTimr emerged. At the same time, the specification expanded from a personal desktop tool into a paid product: Supabase magic-link accounts, Stripe subscriptions, a seven-day trial, backend AI processing, Electron login callbacks, secure token storage, and cached offline access.
What it taught me: turning a personal utility into a product is an architecture change, not a pricing toggle. Identity, billing, secrets, offline behavior, and failure states all become part of the experience.
Saved naming discussion + implementation specification / features specified, not all confirmed shippedThe specification reportedly became working infrastructure.
A later development update reported that authentication, billing, and secure backend API work had been implemented, with builds and type checks passing. Deployment work continued afterward.
What it taught me: passing builds and type checks means the architecture can stand up. It does not yet prove that an installed product survives the ugly shit outside development mode.
Implementation reported in retained conversation / not independently retestedThe packaged app fought back.
The installed macOS build hit a “Fatal JavaScript out of memory” error even though development mode reportedly ran. Authentication links also opened a generic or incorrectly associated Electron application instead of the intended FlowTimr install. The work moved through startup behavior, multiple instances, packaging, macOS URL ownership, removal of remaining Andris branding, and the move to the
flowtimr://callback scheme.What it taught me: “it runs on my machine” is not a product milestone. Installation, startup, identity, and deep-link routing are part of the product whether you planned for them or not.
Reported errors + troubleshooting / permanent resolution not assumedLocal time learned how to travel.
A development update reported local SQLite storage connected to cloud synchronization, including push and pull operations, deletion handling, and a visible sync interface. The migration covered clients, projects, time entries, timer sessions, and app settings, with access rules separating each user’s data. A later update reported a successful production deployment after earlier failures.
What it taught me: once work can move between devices, sync is no longer a feature bolted onto a timer. It becomes the trust layer: offline records, conflicts, deletion, ownership, and recovery all have to make sense.
Implementation + production deployment reported / supported by later architecture specificationThe menu bar became the product’s pulse.
The toolbar icon was designed as its own tiny, bold asset instead of a shrunken app icon. Four states were defined—idle, recording, paused, and needing a prompt—with inactive white or gray, active lime, possible animation, configurable colors, and a popover for the active task and stop control.
What it taught me: the best timer interface might be the one that almost disappears. FlowTimr needed to remain available without dragging attention away from the work it was supposed to remember.
Design decisions + implementation prompt / every animation option not confirmed shippedThe website stopped explaining and started demonstrating.
The recovered prototype introduced the line: “You stay in flow. Timr handles the rest.” It recreated a Mac-style desktop with a menu bar, dock, movable app window, animated timer, and a guided demonstration built from real FlowTimr screenshots and video.
What it taught me: an unfamiliar product is easier to understand when the website behaves like the product. The demonstration itself became part of the interface language later used in the public launch.
Saved interactive website prototype + supplied product mediaThe library came before the editor.
The July project book deliberately pulled FlowClp / VAULT back from becoming a full professional editor. The first product was defined as a local footage library built to answer four questions: What is this clip? Is it useful? Where does it belong? Have I already used it?
Import files, folders + watched foldersThumbnails, proxies + rapid scrubbingHearts, ratings + review queuesTags, projects, collections + notesUsed, posted + publication historyMetadata, transcript + natural-language searchOptional AI descriptions + suggested tagsFingerprints, duplicates + relinkingReveal, open, copy selects + export listsThe defining distinction was usage history, not favorites. A clip could belong to several projects and remember where it had already appeared—supporting a request as specific as: “Show me vertical waterfall footage from Glacier that I have not already posted.”
The editing work was preserved rather than deleted: timelines, audio, trimming, waveforms, snapping, crossfades, merged export, and beat synchronization could return later as a separate Create or Auto Edit workspace.
What it taught me: cutting scope is not abandoning the bigger vision. The useful first product was footage memory. Editing could remain an expansion instead of swallowing the library before it had proved itself.
July project-book specification / high-priority scope, not proof every capability shippedThe three apps started becoming one system.
Shared infrastructure became a defined Flow Platform project instead of something FlowTimr, FlowPic, and FlowClp would rebuild independently. The plan covered a common account, subscriptions and access rights, device sessions, account settings, and a synchronization format for local records.
FlowTimr was designated the reference application. A second app would be the actual reuse test. FlowPic and FlowClp could then share lower-level packages for folders, fingerprints, caches, tags, projects, collections, search, and background processing while keeping interfaces specialized for photos and video.
What it taught me: shared code is only a platform after a second product survives it. FlowTimr could establish the foundation; another Flow app had to prove it was reusable instead of merely familiar.
Shared-platform specification / plan, not evidence one account already worked across every appMarket signal mapped.
The first retained competitive research established the practical battlefield: Mac-native automatic tracking and AI summaries were becoming expected, while invoice/export handoff, post-hoc editing, privacy, and human-approved sharing remained useful openings.
Research run / 22 sources / 6 findingsThe product found its sentence.
FlowTimr stopped being described as another timer and became work memory for independent professionals. The promise sharpened into: “Your work happened. FlowTimr remembers it.”
Positioning dossier / private sourceThe origin became a case study.
NDRS published the first FlowTimr product case study: a contracted employee-time solution that evolved into private work memory for freelancers, independent professionals, and people with ADHD.
First public NDRS product case studyThe website became the demo.
The public marketing site was replaced with an interactive 16-chapter product experience. The guided tour was then stripped of its visible control bar and made to run automatically from the first scene to the trial screen.
Interactive v1 + autoplay v2 / desktop and mobile QADistribution became intentional.
The free-trial flow began collecting a real email before revealing the product paths: FlowTimr 1.0.12 for Apple silicon, the web application, or the smaller Intel build for old Apple shit that still refuses to die.
1.0.12 Mac builds / real lead capture / reversible release
RECOVERED HISTORY / MAJOR TURNING POINTS
The moments where each product became itself.
From timer to installable product.
Accounts, subscriptions, installation, and synchronized records changed the problem from tracking time on one machine to operating a product people could trust.
From folders to meaning.
The archive became searchable by what an image contained and adaptable through reusable organization recipes—not only by where a file happened to live.
Library first. Editor later.
Audio and automatic-editing work expanded, then was deliberately separated from the footage library’s first release so the core memory problem could be solved cleanly.
Three apps became a platform question.
Identity, billing, synchronization, and reusable media infrastructure became shared systems—with the warning that shared code is not proven until a second app survives it.
FlowTimr captures what you worked on. FlowPic helps you find the image you remember. FlowClp helps you find footage worth using—and remember where you already used it.
Evidence matters. The 2025–July 2026 entries are reconstructed from older conversations, the project book, saved specifications, reported implementation updates, and retained prototype material supplied by Andris. Each entry says whether it represents a concept, specification, design decision, reported implementation, troubleshooting, or a retained artifact. The September 2026 entries are backed by retained research, published pages, releases, and QA artifacts. The original VAULT prototype date, a verified FlowPic build history, completion results for several specified editing features, and development after the July 29 snapshot remain unknown until dated evidence is recovered.