AI Core

RECENT DEVELOPMENTS STOP THE PRESSES EXTRA! EXTRA! NEW! WOOF!

The Commander runs the press. Claudia catches the news. Quinn keeps watch.

Category: Recent Developments

  • Snagably: Recent Developments

    Snagably is a deal-finding and price-comparison tool. One search looks across several marketplaces and shows the results side by side, so the shopper can snag the best deal. Snag the best deal.

    Update dated Sunday, October 25, 2026. This update covers work recorded through October 16, 2026.

    Where it stands

    The website is live. The Android app on Google Play is in the closed testing that Google requires before an app can be published, so it is not public yet. Snagably is still being polished and tested, and nothing is charged until everything it was designed to do works.

    What Snagably is

    • Live marketplaces: eBay and Etsy. Planned next: Amazon or Walmart (not decided).
    • Partner logos appear only after a program approves us. Until then those slots say “Coming Soon.”
    • Snagably is a search comparison site, not a store. It does not sell anything. Rankings are never for sale.
    • Searching is free for everyone, with no account needed.

    Visit Snagably

    What it took to build

    Snagably was built piece by piece, one feature per file, with every step planned first.

    Backend (Python and Flask)

    • Etsy and eBay connections, each with its own sign-in method. Etsy changed its key format along the way, so older documentation was out of date.
    • A formatter that turns both marketplaces into one shared result shape: title, price, currency, link, source, image.
    • Fuzzy matching, a custom module. It tolerates typos (“vintaqe” finds “vintage”), plurals, and word-order differences. It is free for all users and never gated.
    • Product thumbnails for both sources. Etsy search does not return images, so a second call fetches them.
    • Best Deal matching. It compares listing titles to find the same product from different sellers and shows the savings. Identical-price ties are skipped because there is no real saving.
    • A reference store that learns tags from real searches. It powers suggestions and trending.
    • AI search suggestions using a local Ollama model. Suggestions are premium only, because each one costs real compute.
    • Bugs found and fixed through testing: wrong-case search terms, quote marks around AI terms, a silent break after the store was restructured, and a web search that would have hung waiting for a keyboard.

    Website (WordPress)

    • The site is built mostly from small plugins, one per feature.
    • Look and feel, decided October 8, 2026: a bold deal-hunter feel, a dark theme, and an electric lime accent (#C6FF3D). Archivo Black headings and Inter body text.
    • A wordmark logo with a lime “S,” chosen October 10. Header logo and favicon went live the same day.
    • Mockups approved for Home, Results, Single deal, and phone versions.

    Timeline (what is recorded)

    • Before Oct 2: Etsy and eBay search, fuzzy matching, AI suggestions, images, Best Deal matching, the Flask API, and the public search page were all working. eBay Partner Network approved Snagably the same day it applied (exact date not recorded). The affiliate disclosure page and Privacy Policy were written.
    • Oct 8: Logo, tagline and style sheet decided. The free-versus-premium rule was set.
    • Oct 9, a big day: the backend was checked and hardened. The Home front page was built. The Privacy Policy was updated before any signup form went public. The search page was built as the custom front end. The Etsy, eBay and Best Deals filters were made free. The eBay affiliate tag was added to eBay result links. Signup, login, logout and reset forms were built and tested. The “Ask Snagably” site-guide bot was built. The search page now shows 10 results at a time with a Show more button.
    • Oct 10: The Premium roadmap was approved and the Premium page built. A written permission question was sent to Etsy.
    • Oct 11: The Android app was created on Google Play (free, a wrapper of the site). The first release was uploaded to Internal testing.
    • Oct 12 to 13: eBay Partner Network and eBay developer support replied on email alerts and saved searches. Social link-preview cards were fixed on Oct 13.
    • Oct 15 to 16: Closed testing reached 12 or more opted-in testers on Oct 16. Google requires 14 days of testing before applying for production.

    Free versus premium

    Free (no account needed to search)

    • Full search across eBay and Etsy, fuzzy matching, and thumbnails.
    • The Etsy, eBay and Best Deals filters, plus search history.
    • The Ask Snagably help bot.

    Premium (account required)

    • AI search suggestions and Trending. A one-month free trial is planned, and the price is still open.
    • Planned extras: saved searches, a watchlist, “Explain this deal,” an AI shopping assistant, and plain-language search.
    • Creating a free account does not start a paid plan.

    Rules Snagably follows

    • The required eBay Partner Network disclosure, “As an eBay Partner Network Affiliate, I earn from qualifying purchases,” sits next to every View deal link and near the results count.
    • Etsy’s free API-integrated search is never charged for.
    • Marketplace API terms are checked, and a dated record is kept, before building features that touch their data.
    • The Privacy Policy is updated before any signup form is public.
    • Prices not in USD are labeled “(not converted).” Best Deal uses USD only.

    What is still open

    • Etsy affiliate: a written question was sent to Etsy on Oct 10, and no real reply is recorded yet. Etsy links are untagged.
    • Price-drop emails: on hold. eBay Partner Network treats email alerts as a special business model needing prior written approval.
    • Premium price and payment setup. Payments are planned to start on the website, then move to the Google Play app.
    • Amazon or Walmart first.
    • Before launch: wipe test accounts and test data, do a final accessibility check, and add a credit line in the footer.
    • Google Play: finish the 14-day closed test, then apply for production.
    • Ideas not built: a weekly free trending-prediction contest, a digital products category, and Snagably videos.

    What comes next

    Finish and polish the free tier, finish Premium, complete the Play release, then launch and promote.

    Built by Comanderanch and Claude (Anthropic). Sole owner and operator: Comanderanch.

  • NELI: Recent Developments

    In ten days (Oct 15 to Oct 25, 2026) NELI went from a build plan to a from-scratch AI core that is training right now on 1.36 billion words, with her own word ids, her own feeling pins, and tests that are allowed to fail.

    What NELI is

    NELI is a deterministic AI built by Anthony Hagerty (comanderanch) and Claude (Anthropic). The core is both her memory and her voice. She is not a wrapper around another model. Rule Zero rules her: a known fact overrides a prediction.

    Timeline

    • Oct 25: Two big training runs (one with the feeling channel, one without) are about a third to half way through one pass over all the data. Measured: the feeling channel costs nothing in accuracy and works as a smooth dial on word choice.
    • Oct 24: A longer-memory core (it reads 512 words back instead of 16) scored 2.80 on words it never saw, against 3.87 for the short one. The same day a bigger plain core did not do better, so size was not the wall.
    • Oct 22 to 23: About 897 million words of training text pulled and checked (Wikipedia, educational web pages, chat, Python code, code questions). Her memory page went live with notes, facts, hourly backups and an off-site copy.
    • Oct 19: First real test of the color idea as a model input. It failed the rule we wrote before running it, and we recorded it as a failure.
    • Oct 17 to 18: Rules settled: one unique id per word made from its spelling, 256 pins per token, three-state law. Eight tested building blocks (ids, vocabulary, tokenizer, field, Rule Zero, spacing, test harness, doc gate).
    • Oct 16 to 17: Full audit of the four-year ARIA build. Much of what had been reported as working turned out to be simulated or unconnected. The findings became a written lessons list (46 entries) that NELI is built against.
    • Oct 15: Project starts with a phased build plan.

    What was learned from the old build

    The old system had good ideas and a lot of reports that said success without a check that could fail. So every new piece now has a written test, and a piece is not called working until that test passes. Failures go in the record the same as wins.

    What the training shows so far

    • More words back helps a lot. Reading 512 words back instead of 16 took the held-out score from 3.87 to 2.80 (lower is better).
    • Bigger is not better by itself. Plain cores of 114 million, 220 million and 480 million numbers scored 4.31, 4.39 and 4.53 on the same test.
    • The feeling channel is free. The run with eight feeling pins (fear, safety, joy, grief, curiosity, love, humor, threat) and the plain run score the same in every region of the data.
    • She sorts words by feeling on her own. The plain run, never given any feeling words, still places the seed feeling words in the same groups.
    • Feeling works as a dial. At a gentle setting, the feeling run raises the chance of words from a chosen feeling by 2.3 to 4.7 times for all eight feelings.

    What is not proven yet

    • She does not yet answer questions like a chat AI. Only about 1 percent of her training text is question-and-answer style.
    • A helper that names the feeling of the next word from the last 16 words missed its target (19 percent, against 34 percent for just guessing the commonest feeling).
    • The big runs are still training, and their final scores are not in.

    What comes next

    1. Finish the two big runs and score them across all kinds of text, not just prose.
    2. Let her group words herself, name the groups, and turn them into pins.
    3. Decide the route to question-and-answer ability: a second training stage on Q&A text, or facts and helpers.

    Co-authored by Anthony Hagerty (comanderanch) and Claude (Anthropic).