Back to home
Build in public

How Orty gets made — mistakes included.

One indie developer, one web app, zero venture money. This log holds the releases, the milestones, the lessons, and the honest mistakes. Newest first.

  1. releaseOct 7, 2026

    v2.1.3 — No key needed: we run the AI now

    The thing that stopped people at the door wasn't the price — it was the setup. “Paste a Gemini key, then a Groq key if you want Co-Pilot” is a sentence that loses a lot of freelancers who just want to know whether a client is worth their time. So v2.1.3 flips the default: Orty now runs a managed GLM-5.3-Flash engine. No key to paste, no provider to sign up for — you log in and score a prospect. For the people who specifically want their own key — cost control, a provider they already trust, or just preference — I added the $5/mo Signal plan: every feature unlimited, your key, and you pay the provider directly with no markup from me. Pro stays $19.99/mo, now $179/yr if you'd rather pay annually, and includes the managed AI. The whole point is to remove setup friction, because a tool nobody finishes configuring is a tool nobody uses.

  2. releaseSep 20, 2026

    v2.1.2 — Orty Scout: the step before every lead

    The most common thing I heard from freelancers wasn't “I can't write the email” — it was “I don't know where to find clients, or which ones are worth pursuing.” So I built Orty Scout. It's now the first thing you see at /Scout. You tell it what you sell and who you want to reach in a few seconds, and it builds a Client Hunting Strategy: the best platforms for your specific target (ranked, not a generic list), platform-specific recipes for YouTube, Instagram, TikTok, LinkedIn, Google operators, Reddit, X and communities, copy-paste queries with one-click Copy and Search, the opportunity signals that make a prospect worth your time, and the negative signals that mean walk away. What it deliberately does NOT do is sell you a giant database it can't stand behind. Scout produces Search Strategies, never invented leads — no fake companies, handles, emails, follower counts, or funding. Found someone? Capture them in seconds, get a fast preliminary 0–100 score, and one click sends them into the full Client Opportunity Report. It runs on Groq, and falls back to a built-in strategy engine if no key is present. Ten relevant prospects beat a hundred generic ones — that's the whole idea.

  3. releaseSep 18, 2026

    v2.1.1 — a Today command center, and the whole UI gets a shadcn pass

    The question Orty answers got sharper: not “what's the score?” but “who should I contact next, and why?” A new Today view builds a prioritized action queue from your own CRM — high-opportunity leads you haven't contacted, replies waiting on you, and follow-ups that are due — with a thin pipeline-health strip, your daily quota, and this week's goal progress. That's pure stats over your entries; no AI call, nothing leaves your account. Under the hood I rebuilt the interface on a shadcn/ui component layer (buttons, tabs, inputs, selects, checkboxes, switches, cards, dialogs, toasts, tables, progress) with a cleaner Orty palette — coral demoted to an accent instead of covering everything, success staying teal, and a near-white default button. The report also gained a confidence level (Low/Medium/High) computed from how well the evidence actually supports the conclusions, not how long the input was. Fixes too: the provider selects stopped clipping their arrow, and the Settings panel is centered again.

  4. releaseSep 15, 2026

    v2.1 — the score leads with money, and Orty opens to every freelancer

    Every prospect now gets a Client Opportunity Score: an overall 0–100 plus Budget, Need, and Fit ratings, an estimated budget band, and the positive/negative signals behind the number — money potential weighted above popularity. You can score a Reddit hiring post, a job listing, a LinkedIn/X profile, or a company site, not just a profile screenshot. Orty widened from video editors to fifteen freelance disciplines (designers, developers, copywriters, marketers, photographers, and more) while keeping editors the default, and a new Discovery Brief hands you where to look, copy-paste search queries, and a scoring checklist — no scraping, no auto-messaging. Rating feedback now feeds the rubric, upgrades return you straight into the app via a pinned Whop checkout, and the old promo-code path is gone.

  5. releaseSep 6, 2026

    v2.0-web — login without leaving, pay without leaving, and a smarter engine

    Orty now has a branded login page powered by Whop OAuth and embedded Whop checkout directly on the pricing page — no redirects to buy. Per-tab deep links (/app/rate, /app/copilot, …) make every screen bookmarkable. The rating engine upgraded to Gemini 3.8 Flash with automatic fallback to 3.7/3.6, and the redesigned 12-question Profile Builder outputs portfolio + rating rules + email rules; a new Wins quick-add panel logs case studies in seconds. We also built an internal rating benchmark harness (tests/benchmark) — numbers to be published once the run completes.

  6. releaseSep 1, 2026

    v2.0-web ships — Orty becomes a web app

    The Windows installer era ends: Orty now runs at getorty.work on Cloud Run (europe-west1), works on Mac/Windows/mobile, and login is Whop OAuth (PKCE). Pro becomes $19.99/mo via Whop — the old $29 one-time was desktop-only and is sunset. CRM moved from local CSV files to Supabase rows synced per whop_user_id (server-side service-role, scoped in the app layer) with a local browser cache; BYOK keys are per-request and never stored server-side long-term. Desktop .exe v1.20.2 stays supported for existing buyers through October 1, 2026.

  7. releaseAug 27, 2026

    v1.20.2 ships — the blank Profile box is gone

    8-step questionnaire replaced the empty editor. Answers become structured JSON sent to Gemini 3.6 Flash — empty achievements are never invented. Added Review with inline Edit + 0–100% strength bar, 5 tones (Professional/Friendly/Confident/Creative/Premium Agency), Copy/Markdown/Plain per section, and local-only persistence (orty_profile_answers_v1). Builder UI lives in Free and Pro; generation stays Pro-gated (ai_profile_gen 403). Improve Existing Profile now also via Gemini.

  8. releaseAug 25, 2026

    v1.20.1 ships — Pro becomes real

    Pro stopped being a label. Login and subscription verification run through Whop OAuth; your API keys never touch our servers. Added server-side gates for batch/video/email/profile/portfolio (Free capped at 1), per-launch localhost token, and export sanitization. Frontend now fail-closes to Free. Unified version + badge fix.

  9. releaseAug 19, 2026

    v1.20-beta ships — the biggest release so far

    Outreach Diagnostics, the guided first-run walkthrough, the 7-stage CRM pipeline, local Personal Insights, weekly goals, one-click workspace export, Trust Settings with View-My-Data, and a Free tier. Eight headline features in one drop — beta buyers keep full access under their original terms.

  10. lessonAugust 2026

    The rate-card guardrail became the brand

    Early on, Co-Pilot was 'an AI that drafts emails.' Users didn't repeat that sentence. What they repeated was: 'it can't invent prices.' Constraints turned out to be more memorable than capabilities — the marketing now leads with what Orty refuses to do.

  11. milestoneAugust 2026

    Launched on Product Hunt as Orty

    The OutreachRater rename deserved a real launch. Product Hunt brought the first wave of strangers using the app end-to-end — which is exactly the stress test a solo-built installer needs. Every bug report went straight to the founder's inbox and got answered personally.

  12. mistakeSummer 2026

    I shipped features before the onboarding could carry them

    v1.x had real power hidden behind a blank Rate tab — and beta users told me they scored one lead and stopped. v1.20's guided walkthrough exists because of that mistake. Lesson: a feature without an onboarding path is a changelog line, not a product.

  13. noteSummer 2026

    Why local-first isn't a marketing angle here

    Orty handles screenshots of real people and draft emails that quote real rates. The cheapest architecture to build would have been a cloud pipeline — and the fastest way to lose this exact audience. Local-first was the product decision that made every trust claim on this site true by default. Addendum (Sep 2026): Orty moved to the web — the principle survived as 'your keys, your data': BYOK, session-only secrets, occurrence-only analytics.

  14. notev1.1-beta

    OutreachRater becomes Orty

    The old name described the mechanism ('rate outreach'). The new name left room for everything the app grew into — diagnostics, CRM, goals. Renaming after early users existed meant migrating their data carefully; the CSV auto-migration shipped in the same breath.

  15. lessonDay one

    Built because my own inbox needed it

    Orty started as a personal script: score which creators were worth emailing before spending an evening writing pitches. It sent 25 real outreach emails in its first week of use — that number on the homepage is the founder's own result, kept honest as a single data point rather than dressed up as an average.

Want the next entry to be something specific — a deep dive on the scoring rubric, real churn numbers, a postmortem? Ask: hello@getorty.work reaches the founder directly.