The roadmap

What works now, what ships next, and what is still rough.

Search comes first. Everything else follows only if the core engine earns trust, so this page starts with launch reality before it shows the wider toolset.

A note before launch

Version one will have edges.

I am one person, and the first public version of Argand will reflect that. The corpus is going to be small compared to anything you are used to. Some searches will come back empty. Services like Maps, Scholar, and Discovery will roll out in stages over the months after launch, not all at once. Some things will work fine in testing and break the moment real people touch them. None of that is a disclaimer; it is just how shipping software goes when one person is doing it.

What I can promise is this:

  • Every email to nic@argand.org goes straight to me, gets read, and gets answered.
  • Every failed search you report gets investigated, because real failures are the only useful guide to real fixes.
  • The engine can improve from reported failures and anonymous aggregate signals without tying analytics to you.
  • When a fix is ready, I can ship it without waiting for a quarterly product cycle.

Thank you for being early. The thing only gets good if you tell me when it is not.

Nic Weyand

Tell me what you'd want from it

An unusual section to put on a launch page

Things that will be rough on day one.

Most pre-launch pages only show the polished parts. This is the list of places Argand will not be ready yet. If you came for one of these and got nothing back, this section is the explanation.

  1. Corpus size.

    The largest indexes are measured in the trillions of pages. At public launch Argand's will be in the millions, with priority coverage of the domains people actually want (open web reference material, academic, news, technical documentation) and weaker coverage of the long-tail commercial web. The gap closes with each crawl.

  2. Breaking news latency.

    A traditional crawl-and-index pipeline lags real-world events by hours. Until PressTrace ships the streaming ingestion path, Argand will show yesterday's news, not the last ten minutes'. For active events, a larger live-news engine still wins.

  3. Image and video search.

    Visual search exists internally but won't ship in version one. Reverse-image lookup, screenshot-to-source, and video keyframe search all come later. If you need to find the original of an image you've got, TinEye or a larger image engine will still be the answer at launch.

  4. Shopping and product search.

    Shopping search is missing by design. Product comparisons and price tracking would need their own anti-bidding system, because every shopping engine I have seen is tempted to sell the click to the highest bidder. That violates the Traffic-Out Principle, so it is not part of version one.

  5. Maps, Discovery, Scholar, the rest.

    Most of the tools on the roadmap are in development, not ready. Search is what ships first; the others roll out in stages over the months after launch as each one clears its own quality gate. The waitlist tells you when each goes live.

  6. Languages outside English / a few EU languages.

    The first index is heavy on English, with reasonable coverage of French, German, Spanish, Italian, and Dutch. Other languages will return results but quality is uneven. The embeddings are weaker, the spell-correction dictionaries smaller, and the gold-set for the regression gate thin. Quality climbs with crawl volume and with community-contributed evaluation queries.

None of these are excuses. They are commitments. Every one is on the roadmap, every one has a build path. If you hit a gap that isn't on this list, that's worth knowing about: nic@argand.org.

Why this is

Mostly: this is one-person work, and the hardware got expensive.

If you read that list and thought "sure, but why?", here is the honest version. It comes down to a few things.

First, this is one-person work. The biggest search engines have thousands of engineers and the budget to match, while most of Argand's missing pieces are simply waiting for me to get to them.

Tap, swipe, or use the arrow keys.

None of this changes much in the next six months. A year after launch most of it will be a little less true than it is at launch. That is about as honest as I can make it.

Prefer to read? The honest version in full.

It is only me, without a cofounder or employees. The biggest search engines have thousands of search engineers and the budget to match. Most of what is not ready yet is not waiting on some hard breakthrough; it is waiting on me to get to it. There are weeks where the most important thing I can do for the project is sleep.

I was planning to scale up the crawl and the index in early 2026. Then everything that involves a GPU got expensive. A consumer card that was around 1,600 dollars in 2023 is closer to 3,500 now, when you can find one. I have a saved tab that refreshes the Nvidia store every few minutes for a card I needed in February. Cloud GPU rentals are up roughly fivefold, because every chatbot startup with a check is renting the same capacity I am. So Argand is running on the index I could afford, not the index I planned for.

The largest engines have been crawling the web since I was a kid. Argand started in 2025. They have had close to three decades to learn how the web behaves at scale, sorting real pages from spam and learning which links rot the day after you index them. Argand has to learn all of that from a cold start. It gets better every crawl. The deep historical stuff that vanished ten years ago probably never fully catches up.

Part of what is missing is missing because I refuse to ship it. There is no tracking pixel telling me which results people lingered on, which is what engagement-based ranking actually is at every other engine. I will not put an AI summary above the results, because if the summary is the answer, then the ranking underneath can be garbage and no one notices. Shopping search also has to wait, because every shopping engine I have ever seen is pulled toward whoever paid the most, and that is exactly what Argand exists not to do.

The Argand toolset

Search is the first job.

Behind Argand are 59 small services and 47 in-house Rust crates. The twelve tools below are the public ones. Each one starts from the same standard: do the job quickly on ordinary hardware, keep the user's life out of the product, and stay answerable to the person using it.

Tap any service to see why it exists and what it does.