Legal · copyright, terms and the privacy note
Written to
be read.
Most pages like this one are written to be skipped. This one is not. It is about this site rather than any site: every statement below describes something that is actually in the code of labs.llc, and can be checked against it. Exactly one paragraph does not work that way — the choice of law in clause 12, which is not the sort of fact code can settle — and it says so where it stands rather than hoping you will not notice.
There are thirteen clauses. Each has its own address, so any one of them can be quoted and linked exactly — the paragraph mark beside a heading copies that address. Nothing here asks you to agree to anything before reading it.
thirteen clauses · every one linkable by its own address · last changed 19 August 2026
01 · Who this is
Who this is, and how to reach us.
¶labs.llc is published by LUXURY AMERICAN BRANDS LLC. The site and every property on it are signed LABS, which is the name the work appears under; the two refer to the same publisher. Where these terms say “we” or “this site”, that is who is meant.
Formal notices go to legal@labs.llc. A rights concern, a takedown request, a question about these terms or a correction to something on the estate all belong there.
If that address ever bounces, use labs@labs.llc, which is the desk printed on every other page of this site and is read daily. Two addresses are named rather than one because a legal page exists so that a notice reaches somebody: naming a second, older, certain route costs a line and removes the one failure that would matter.
- Legal notices
- legal@labs.llc — a rights concern, a takedown request, a question about these terms, a request to remove something, or any formal notice. Clause 11 says what to put in a rights message.
- Everything else, and anything that bounces
- labs@labs.llc — the general desk, and the address printed on every other page of this site. Corrections, questions, bug reports, a figure that looks wrong, a claim on this very page that does not match the code. Nothing is lost by writing here instead; put Legal at the front of the subject line and it is read as a notice.
A person reads it, not a queue. Please never send a card number, an account number or a government identity number to either address — there is no reason this site would need any of them, and a page that asks you for one is not this page.
02 · Copyright
Copyright, and what is not claimed.
¶The text on this site, the code that runs it, the design system it is built on, the drawings and product plates made for it, and the way each desk is arranged, are our work. Copyright © 2026 LUXURY AMERICAN BRANDS LLC. All rights reserved, except as clause 4 gives them back.
Being exact about the boundary matters more than the claim, because a great deal of what appears on these pages is not ours and we do not pretend otherwise. No copyright is claimed in any of the following:
- The public data the live desks report. A magnitude from the USGS, a tide height from NOAA, a rule from the Federal Register, a train from the MTA, a debt figure from Treasury — the numbers belong to the bodies that published them, and reporting a number does not create a right in it. Clause 9 names the sources.
- Other people’s names and marks. Where a directory names a restaurant, a jeweller, a hotel, a maison, a shoe or a card issuer, that name is theirs. See clause 3.
- Material carried under someone else’s licence. Some data arrives with terms attached and those terms travel with it: the padel register is an OpenStreetMap extract under the Open Database Licence, credited to OpenStreetMap contributors; the language register is Glottolog 5.3 under CC BY 4.0, with its citation printed on the page; Wikimedia material is under Wikimedia’s own licences. Those licences govern that material, not this clause.
- Third-party libraries. Where a page loads one it stays under its own licence, and there are more of them than a single example would suggest: the FFmpeg build OmniFile fetches and the ZIP library it ships, Tailwind, Font Awesome, Tone.js, three.js, Chart.js, html2canvas, jsmediatags, Leaflet, chess.js and a QR encoder. Clause 8 says which of them arrive from somebody else’s server while you read.
What is left after all of that is the writing, the code and the design, and that is what this clause covers.
03 · Trademarks
Marks, ours and other people’s.
¶Ours. The names labs.llc and LABS, the wordmark, and the names of the properties published here — among them Songbuilder, OmniFile, Watermark, File Share, Glass, Papertrail, Yum, Newsy, DomainChaos and Product Laboratory — are used as marks of this publisher. Please do not use them in a way that suggests we made, endorsed or ran something we did not.
Everyone else’s. Every other name, logo and mark on this site belongs to whoever owns it. They appear here to identify the thing being written about and for no other reason — a restaurant in a guide, a jeweller in the 47th Street directory, a hotel in a city book, a maison, a shoe, a bank, a card issuer, a news publisher whose headline is on a wire, a data source named so a reader can go and check it. That is nominative use: naming a thing in order to talk about it accurately.
We are not affiliated with any of them. Nobody named on this site sponsors it, endorses it, reviews it, approves it or has been asked to. Nobody pays for placement anywhere on this site, no listing has ever been sold, and there is no commission, referral fee or advertising on any page. If a business would rather not be listed at all, write to legal@labs.llc and it comes off — no reason required, and no argument.
04 · Permitted use
What you may do.
¶A page that forbids everything gets ignored, and rightly. Here is what is welcome, stated plainly enough that you should not need to ask.
- Read any of it, for anything. Nothing here is behind a paywall, a registration or a reading limit, and nothing asks you to make an account in order to read it. Personal use, professional use, classroom use and commercial reading are all the same use as far as this site is concerned.
- Link to any page, from anywhere. Deep links are fine and are meant to work: every clause on this page, every section of every desk and every plate has its own address. You do not need permission to link and you never have to ask.
- Quote it, with attribution. Take a paragraph or a figure for a review, an article, a paper, a lesson, a thread or a brief, and say where it came from — naming labs.llc and linking the page is enough. Fair use and fair dealing are unaffected by anything on this page.
- Use the browser tools for your own work. Whatever you make with Songbuilder, OmniFile, Watermark, Glass, Bench, Guitars, Drums, the Tuner, File Search or the game boards is yours. We claim nothing in your output, your recordings, your converted files or your text, and we take no licence in them — which is easy to promise, because for most of those tools your file never reaches us at all. Clause 8 says which ones and why.
- Take a screenshot, and print the page. Both are fine, in any context, including a commercial one.
- Read the source. The pages are plain HTML, CSS and JavaScript with the reasoning written into the comments. Read them, learn from them, and check any claim on this page against them — the last of those is why they read the way they do.
05 · What we ask
What we ask you not to do.
¶Four requests. Each has a reason, and the reason is the point — if you can meet the reason another way, that is fine by us.
- Please do not fetch so hard that the site stops answering. Reading it with a script is fine. Hammering it is not, because this is a small site on ordinary hosting and a fast enough loop takes it down for everyone else reading at the time. Go at a human pace, honour a 429, identify yourself in your user-agent, and if you want a lot of it, write and ask — the answer is usually a copy of the file.
- Please do not republish a directory wholesale. Quoting entries, citing figures and linking rows are all covered by clause 4. Lifting a whole register — the restaurant guide, the 47th Street directory, the plate archive, the city books — and re-hosting it as your own listing is different: the value in those is the compiling and the auditing, which took a long time and is the part that is actually ours.
- Please do not pass the work off as your own. Republishing a page or a plate without attribution, or feeding it out under someone else’s byline, is the one thing here that is simply bad faith. Attribution costs a line.
- Please do not drive the relays from somewhere else. A few endpoints on this site exist so that this site’s own pages can read a source that refuses a browser directly — the certificate ledger’s relay, the power desk’s, the flight desk’s, the storm desk’s and the news feed reader. Four of the five can reach exactly one upstream, written into the file; the news feed reader works from a list of permitted hosts instead, because it reads many publishers. All of them validate the parameter you send, resolve the host once and connect to that exact address, and refuse anything that resolves to a private or loopback range. What they do not do is check who is calling. There is no same-origin test and no per-caller rate limit in any of them, so nothing but this paragraph stops you pointing your own application at one — which turns a small courtesy into an open proxy, and is how such things get switched off for everybody. Call the upstream yourself; it is named on every desk.
None of that is a threat and none of it is new law: it is the same request any small site would make, written down so nobody has to guess.
06 · The data
The data, and what it is not.
¶The live desks read public sources and report what they say. They do not model, forecast, rank, score or advise, and where a desk shows a figure it names the body that issued it and links back to it. That body is the authority on its own figure; this site is a reader of it, and where the two disagree the issuer is right.
Four cases are worth stating outright, because they are the ones where mistaking a reading desk for an authority could actually cost somebody something.
Storm, tide, river and quake — not a safety service
¶The hurricane, tide, river, seismic and space-weather desks are reading desks. They issue no watch, no warning, no evacuation notice, no forecast track and no cone; they draw no navigational chart and carry no soundings. A tide figure for a subordinate station is an offset applied to a reference gauge some distance away, not a measurement taken where you are standing. Do not use any of it for navigation, and do not use it to decide whether to move, to board up or to leave. For that, read the National Hurricane Center, your National Weather Service office, the relevant harbour authority and the official chart, and do what they say. The desks link each of them.
The card desk — not financial advice
¶The credit-card desk reports publicly advertised offers with the date they were last checked printed on the page. Rates, fees, bonuses and promotional windows move constantly, and an offer can change or end between a check and your reading it. Nothing on that desk is financial advice, an application, an offer of credit, or a recommendation, and nobody there is a licensed adviser. Confirm the current terms on the issuer’s own page before you apply; that page is the only one that counts. No issuer pays for a position and there is no affiliate or referral arrangement anywhere on it.
Markets, money and crypto — not advice either
¶The same holds for the market tape, the treasury desk, the currency board and the crypto pages. They are read from public quote and statistics endpoints, they can lag, and they can be wrong when the source is wrong. None of it is investment advice or a recommendation to buy or sell anything.
The directories — a directory, not an endorsement
¶The restaurant guide, the 47th Street jewellery directory, the city books and the padel register are directories. Appearing in one means an entry survived a check, not that anyone recommends the business, and the ordering inside a section is alphabetical or by the source’s own field precisely because any other order would be an opinion this site has not earned. They are not verified registers. A row is present because a check did not disprove it, not because a person confirmed it; entries carry no ratings and no “verified” badges for that reason. Trade turns over, businesses close and numbers change — ring before you go.
07 · Accuracy
Accuracy, and how to correct us.
¶The standard this site holds itself to is narrow and worth stating because it is the whole of the editorial policy: a page may say only what can be stood behind. A figure names its source. A field that cannot be checked is not carried at all rather than carried with a plausible value in it. Where a source says nothing, the page says nothing rather than filling the gap.
Errors still happen, and the honest thing is to expect them and to publish what was found when they were. The audit record is part of this site rather than hidden from it:
- The restaurant guide, audited August 2026. 270 entries did not survive it. 103 were placeholders — rooms literally named “Munich Fine Dining 1” — and 167 were invented variants of a real restaurant, each carrying its own made-up address, rating and review count. All of them are gone. So are the fabricated fields on the rooms that were real: the websites had been generated by truncating the name and 55% of them did not resolve, and the ratings, review counts, star quotes, phone numbers and opening hours had never been checked by anyone. What is left is the room, where it is, what it cooks and its Michelin stars as last published.
- The 47th Street directory. Compiled from a listings dump and audited structurally, with entries checked by hand against the open web. The source carried 1,410 rows, of which roughly two thirds were removed as machine-generated, uncorroborated, or belonging to a market whose sample failed. What is left is what survived.
The rule those two audits produced is the rule now: an entry that cannot be supported is removed, not softened. That applies to anything on this site, including anything a reader points at.
Found a mistake? Write to labs@labs.llc — a wrong figure, a business that has closed or moved, a name spelled badly, a source that has been superseded, a claim on this very page that does not match the code. A correction with a source attached is acted on fastest. A request to be removed is acted on without one.
08 · Privacy
What is set, what is measured, what stays.
¶There is no advertising network on this site, no profile built out of what you read, and nothing gathered here is sold to anybody. Nothing on this site asks you to sign up in order to read it. One property offers an account and it is named rather than glossed: the chat rooms keep an optional register, described under the places where something does leave. A room seats you from its code alone; the account exists so a name stays yours across devices.
Nothing below is a promise about a future version. It is a description of the code that is running now, and every sentence in it can be checked against that code — which is the standard clause 7 sets, applied to the clause readers have the most reason to doubt.
What this site stores in your browser
¶This is not a complete list, and a page that offered you one would be wrong by the next build. A grep for literal key names across this site finds more than a hundred, and a few of those are prefixes that fan out into more. What follows is instead the honest shape of it: the two that follow you everywhere, then the three the menu keeps, then what a desk keeps for itself, and last the ones that single your browser out — because those are the ones worth knowing about.
Two of them are on nearly every page, because they are how the site knows which edition you are reading.
- labs_edition
-
A cookie holding one word —
paperordark— so that a page built on the server knows which edition to render before it writes its first link. Path/,max-age=31536000, which is one year,SameSite=Lax, andSecureover https. It is not marked HttpOnly, because the page’s own pre-paint script reads it too, and because it says which colour you are reading and nothing else. - labs.mode
-
The same choice again, in
localStorage, for the pages that are plain files rather than server-rendered. The value islightordark. Local storage has no expiry: it stays until you clear it.
Three more belong to the full menu, the one with the marks and the shelf in it. Only nine pages here carry the script that writes them — the homepage in both editions, the strip board, and the weather, news, portfolio, investor and crypto fronts. Every other page, this one included, has a menu that writes nothing.
- labs.book
-
One word,
openorshut, remembering whether you left the drawer’s page section folded up. - labs.recents
labs.favs -
The recent and favourite chips in that drawer, in
localStorage, capped at 6 and 24 entries. Where the page you clicked is one of this site’s own registered properties, what is stored is the property — its name in the register — so the list still points at the right edition when you switch. Where it is not, what is stored is the page’s address with anything after a#removed, plus the words on the link. Links to other people’s sites are not recorded at all. The list is built and read in your browser and never sent anywhere.
Then each desk keeps its own handful — the one or two
choices it would be irritating to make twice.
quake.window holds which time window you were reading,
weather.city and weather.unit the place and
the units, hurricane.basin the basin,
transit.station the station, rivers.gauge
the gauge, flights.city and flights.radius
the centre and the range, radio.vol the volume,
money.base the base currency; and so on through the
studio’s sb.* settings, the chat rooms’
mb.style.* ones, the media wall’s page size and
each toast switch. They are readable in your browser’s
developer tools, and none of them is transmitted.
Three of them do single your browser out, and it is worth saying
so plainly. The chat rooms, the card table and the game boards
each mint a random string the first time you open them —
mb.cid, labcards.cid,
labchess.cid and its checkers and noughts-and-crosses
twins — and keep it for good. It is a number your own browser
invented, with no name and no address in it, and it exists so a
live game or a room can tell one seat from another across a reload.
It is sent to this site with every move you make in a shared game or
a room. Alongside it, the chat rooms store mb.seats:
for each room you are in, the seat token and, for a sealed room,
the room’s encryption key — which is the whole
reason a sealed room stays sealed from this site, and also the reason
that clearing your browser storage loses those rooms for good. A
board and a room also keep the display name you typed, because you
typed it.
Three other stores are in use besides local storage, and they are
not the same thing. A few pages use
sessionStorage, which the browser throws away when the
tab closes. File Search keeps the folder you pinned in
IndexedDB — a permission handle, not your files
and not their contents — so it can offer the same folder next
time. The chat rooms register a service worker, which keeps a
Cache Storage copy of that page’s own script,
stylesheet and logo so it opens fast and works offline; messages are
never cached there.
Finally, some cookies exist that almost certainly are not for you.
Signing in as an administrator starts a PHP session, and there are
four separate ones across the site —
labsadmin for the site tools, card_admin
for the card desk, fs_admin for File Share’s
console, and the PHP default PHPSESSID on the
domain-sales page. The first and the last are set HttpOnly and,
over https, Secure, in their own code; the middle two set no flags
of their own and take whatever the server’s PHP configuration
gives them, which this page cannot see and so will not describe.
A signed-in operator of one of the cross-site tools additionally
carries a signed lab_auth cookie for thirty days.
If you have never been asked for a password here, none of these
has ever been set in your browser.
What is measured
¶Nearly every page carries Google Analytics 4 under the measurement id G-Z2N672WDTY — at the last count, 175 of the 184 HTML pages in the build, the nine without it being older duplicates and one verification file that were never given the tag. It is the only measurement of any kind on this site: there is no advertising pixel, no session recorder, no heatmap and no second analytics product anywhere in the build.
What the tag is asked to do is exactly one thing. Every page calls
gtag('js', …) and
gtag('config', …) and nothing else — there
is no custom event, no user identifier, and nothing is ever pushed
into its queue by hand. So what is reported is the default page view:
the address of the page you opened, the page you came from, and the
standard device and locale facts Google’s own tag collects. No
page content and no search term is passed to it, and nothing you type
into a field is read out of that field and sent. Google sets its own
cookies to do this and those are Google’s: what they are, how
long they last and how the data is handled is described in
Google’s own policy,
and a browser-level opt-out is available from
Google.
“The address of the page” has a consequence this page
will not soften. An address includes everything the address
carries, and that includes the part after a #. File
Share’s optional in-browser encryption puts the decryption key
in exactly that part of the link, and the page that opens an
encrypted share — fileshare/f.php —
carries the analytics tag like every other page. So on the
code as it stands, opening an encrypted share link sends that link,
key and all, to Google as the page address, in the ordinary page
view. That is a defect rather than a design, and the fix is one line
in that file; until it is made, this is the true position and you
should have it. Treat an encrypted share link as the secret
itself, send it by a channel you would trust with the file, and
do not paste it anywhere you would not paste the file.
What other companies see while a page loads
¶A page does not have to send anything of yours for somebody else to see you arrive. Where a page loads a font, a stylesheet or a script from another company’s server, your browser asks that server for the file, and that request carries your IP address, your user-agent and the page you were on. It happens on load, before you press anything.
This site does that from seven other hosts, and they are worth naming because none of them is visible on the page. Google Fonts serves a typeface to about thirty-seven pages, the homepage among them. cdn.tailwindcss.com serves the stylesheet engine to sixteen older pages. cdnjs.cloudflare.com serves Font Awesome, Tone.js, three.js and html2canvas; cdn.jsdelivr.net serves Chart.js and a media tag reader; unpkg.com serves Leaflet to the weather map and, on demand, the FFmpeg engine to OmniFile. The crypto pages load around a thousand coin logos from cryptologos.cc and s2.coinmarketcap.com. None of those is an analytics product and none of them is asked to measure anything, but each one sees the request, and a page that described only its own analytics tag while seven other hosts were served the same visit would be telling you half of it.
The tools that never send your file anywhere
¶
Several properties here work on a file you choose, and they do it
entirely inside the tab. This is verifiable rather than asserted: the
following carry no fetch, no
XMLHttpRequest, no WebSocket, no beacon and
no upload target of any kind —
Bench (audio analysis),
Guitars, Drums and the Tuner,
Glass, Watermark (invisible characters in text) and
File Search. Nothing you open in them is uploaded, because
there is nowhere for it to be uploaded to.
OmniFile is the one that needs a fuller answer than “nothing is uploaded”, and it deserves one. Images and text convert with nothing but the page. Audio and video need FFmpeg, which is not part of the page, and pressing Fetch the engine downloads it from unpkg.com: the FFmpeg wrapper and its worker, plus a core of roughly 32 MB uncompressed, served at about 10 MB over the network. That is a large transfer, and it is worth knowing about before you press the button — but the direction is inward. It is compiled code arriving in your tab, and it lives there only until the page is closed. No file of yours goes out, then or ever.
Songbuilder makes its sound in the tab and no audio ever leaves it, including a recording, which is saved by your own browser. It has one deliberate exception: pressing Keep a link posts the song’s save code — a short piece of text describing the arrangement, with its name, style, tempo and key — to this site, which stores it so the short link opens the song. That happens only when you press that button, and only that text is sent.
The places where something does leave
¶Five, and each of them is the point of the thing rather than a side effect.
- File Share exists to share files, so it uploads them. The file goes from your browser to this site’s own endpoint and on to Litterbox, a temporary host, where it is deleted on the schedule you chose — one, twelve, twenty-four or seventy-two hours. What stays here is the small mapping that makes the short link short. A share can optionally be locked with a password, set to self-destruct after its first download, or encrypted in your browser before it goes, in which case the key travels only in the link. That last option is the only one under which the host cannot read the file; without it, treat the upload as you would any temporary host. Litterbox’s own terms apply to what is stored there. Read what is measured before you send an encrypted link anywhere.
- The chat rooms carry messages and files, so both are sent. What you type in a room goes to this site and is held for the room; a room created as sealed is end-to-end encrypted for text, which means the copy stored here is ciphertext and the key never reaches this site. A room also accepts file uploads of up to 1 GB, which are stored on this server beside the room. A room and its files are swept thirty days after the last word in it, or ninety if the host turned the keep dial up. If you switch notifications on, the browser’s push subscription is stored here so the server can tickle it — the push itself carries no words. And the optional account: a username, a password stored only as a hash, an activity log the profile shows you, a shelf of transcripts you chose to save, and, if you add one, an email address, which is kept so a forgotten password can be reset by a code sent to it. An account can be deleted from the profile, password-confirmed.
- An online game sends your moves and your name. Chess, checkers, noughts-and-crosses and the card table each have a small relay on this site so two browsers can share a board. What it holds is the move list, the display name each side typed, and the random client id described above — enough to know whose turn it is. A game is swept from the server twenty-four hours after it was last touched. Playing pass-and-play against the machine sends nothing at all.
- The contact form. It takes your name, email address, optional company and phone, the topic, optional budget and timeline, your message, and whether you ticked the updates box. It records with those the page you sent it from, your IP address and your browser’s user-agent string, so that abuse can be traced and rate-limited. The record is written to a directory that is closed to the web, and mailed to labs@labs.llc with the reply address set to yours. The newsletter form takes the email address alone, with the date, and every mailing carries a working unsubscribe link.
-
The desks fetch from their sources. A live desk reads its
source from your browser, which means that source sees the request
— your IP address and your user-agent — exactly as if
you had visited it yourself. The Search desk is the most
explicit case: what you type there is sent to the services it
names in order to be answered, and clause
9 names them. The network and
domain desks are the most personal: to tell you about your
own connection they ask api.ipify.org for your public IP,
ipapi.co where that IP is, and dns.google to
resolve the names you look up — so every hostname you type
into those desks goes to Google. Where a source refuses to answer
a browser directly, a page either uses this site’s own
narrow relay or, on a number of older pages, a public CORS relay
(
api.rss2json.com,api.allorigins.win,corsproxy.io,api.codetabs.com,proxy.corsfix.com,proxy.cors.sh) — in which case that third party sees the request and the address being fetched.
The menu’s own search sends nothing — it runs
entirely against an index file already in your browser, and the file
that does it contains no fetch at all. The drawer around
it is not quite so quiet: the shelf of nine books at the bottom of the
homepage drawer asks Open Library for them each time the drawer
opens, so opening the menu on that page is a request to Open Library.
Nothing about you is sent with it beyond the request itself.
09 · Elsewhere
Leaving this site.
¶This site links out constantly and on purpose: every figure names its issuer so you can go and check it. Once you follow one of those links you are on somebody else’s site, under their terms and their privacy policy, and we have no control over what is there or whether it is still correct. A link is not an endorsement.
The sources fall into a few groups, and what is true of one group is not true of another, so it is worth naming them separately rather than making a single claim about all of them.
- United States federal material. The seismic and river data from the USGS; tides and currents, water levels, space weather and satellite imagery from NOAA; storm advisories from the National Hurricane Center and forecasts from the National Weather Service; rules from the Federal Register, with Regulations.gov and GovInfo; figures from Treasury FiscalData; recalls from the FDA, the Consumer Product Safety Commission and USDA FSIS; historical newspapers from the Library of Congress. A work prepared by an officer or employee of the United States Government as part of their official duties is generally not subject to copyright protection in the United States, which is why this material can be reported here freely. That is a statement about the general position and not about every item: each agency states its own terms, and some pages carry third-party material under separate rights.
- Other public bodies. Transit feeds from the Metropolitan Transportation Authority. These are public but sit under their own developer terms rather than under the federal rule above.
- Open data under a named licence. OpenStreetMap, under the Open Database Licence, credited to OpenStreetMap contributors. Glottolog 5.3, under CC BY 4.0, cited in full on the page that uses it. Wikimedia, whose article text is generally CC BY-SA and whose media items each carry their own licence. Open Library and the Internet Archive. Where a licence requires attribution or share-alike, that obligation travels with the data and is not displaced by anything on this page.
- Public interfaces run by other people. Open-Meteo for weather, Frankfurter for exchange rates, energy-charts.info and the UK’s carbonintensity.org.uk for power, crt.sh for certificate transparency, adsb.lol for aircraft, The Space Devs and wheretheiss.at for launches and orbits, Radio Browser for stations, Datamuse and dictionaryapi.dev for words. Each has its own terms and its own limits, and each is named on the desk that uses it.
- Domains and networks. The domain desk reads IANA’s RDAP bootstrap and then the registries’ own RDAP servers directly. Two older pages — the network tool and the newsroom probe — still go through rdap.org, which is a third party sitting in the middle of that lookup. Those same pages use api.ipify.org to learn your public IP address, ipapi.co to place an IP, and dns.google to resolve names, which means a hostname typed into them is a query sent to Google. Clause 8 says the same thing from the reader’s side, because it is the sort of fact that belongs in both places.
- What the Search desk asks on your behalf. A search there is federated, and the words you typed go to every service it queries: Wikipedia and Wikinews, Wikimedia Commons, api.gdeltproject.org for news coverage, GitHub, OpenAlex for papers, Openverse for openly licensed media, Stack Exchange, arXiv, Hacker News through hn.algolia.com, Lemmy and Mastodon. The desk also offers to hand a question on to an assistant of your choosing, which is a link you press rather than a request it makes.
- Crypto. The crypto pages read more sources than any other desk on the site, and are the one place a WebSocket stays open while you watch: CoinGecko, Binance (the live price stream), CoinPaprika, CoinLore, alternative.me for the sentiment index, and, for chain data, mempool.space, Blockchair, TronGrid, publicnode’s Ethereum and Solana endpoints, drpc and Solana’s own mainnet endpoint. Headlines there come from CoinDesk, Cointelegraph, Decrypt and CryptoSlate. Nothing on those pages is advice — clause 6 — and no wallet, key or address of yours is asked for or held.
- Commercial services under their own terms. Ticketmaster’s Discovery API for the concert bill, Yahoo Finance endpoints for the market tape, Litterbox for File Share’s temporary hosting, unpkg for OmniFile’s engine and the weather map’s, and news publishers’ own RSS feeds for the wires — alongside Google News RSS searches, which are an aggregator’s index of other people’s headlines rather than a publisher’s own feed. Headlines on a wire are the publisher’s; each links straight back to them, and each remains under their terms.
If you are going to reuse anything that came through this site from one of those sources, go to the source and read what it says. That is the only reliable answer, and it is why every desk links back.
10 · Availability
Provided as it stands.
¶This site is provided as it is. There is no uptime promise, no service level, and no undertaking that any page, desk or tool will be available at any particular moment, keep working the way it does today, or continue to exist. A desk depends on a source that may change its interface, throttle it, charge for it or withdraw it without notice, and when that happens the desk stops working until it is fixed.
No warranty of any kind is given: not that the site is free of errors, not that it is fit for any particular purpose, and not that the data it reports is accurate, current or complete. Clause 7 describes the standard we hold ourselves to, and it is a standard rather than a guarantee.
On liability, in plain English. Use your own judgement about anything that matters. To the fullest extent the law allows, we are not liable for loss or damage arising from your use of this site or from anything you relied on here — including lost profit, lost data, lost time, or a decision taken on the strength of a figure that turned out to be wrong. Some jurisdictions do not allow certain exclusions, and where that is so, this clause applies only as far as it lawfully can and nothing in it removes a right you cannot lawfully be asked to give up. Nothing here excludes liability for fraud, or for anything else that cannot be excluded by law.
11 · Rights
If something here is yours.
¶If you believe something on this site infringes your copyright, misuses your trademark, or should not be here for some other reason, write to legal@labs.llc, or to labs@labs.llc if that bounces. You do not need a lawyer to do this and you will not get a form letter back.
Please include enough for the material to be found and the claim understood without a second exchange:
- The work
- What it is, and where it can be seen if it is published elsewhere.
- The URL
- The exact address on this site where it appears — a page, a plate, a row in a register, a line in a wire.
- The basis
- Why you hold the right, or that you are authorised to act for whoever does.
- Your details
- A name and an address we can reply to.
- Good faith
- A statement that you believe in good faith that the use is not authorised by the rights holder, their agent, or the law, and that what you have told us is accurate.
What happens then. The message is read by a person, not a queue. Material that is plainly someone else’s comes down while the question is looked at rather than after — the standing rule on this site is that an entry that cannot be supported is removed rather than defended, and that applies here too. You get a reply saying what was done. If we think the use was fair or the material is not yours, we will say so plainly and explain why, and you keep every remedy you would otherwise have had.
A note on process, so nothing is implied that is not true. This site does not claim to have a designated agent registered with the United States Copyright Office, and you should not read anything above as a representation that a statutory notice-and-takedown procedure has been formally established here. The route described in this clause is the practical one: it is read quickly, acted on quickly, and is the fastest way to have something removed. If you need to serve a formal notice, say so in your message and we will tell you where to send it.
12 · Governing law
Governing law.
¶These terms are governed by the law of the State of New Jersey, and the state and federal courts of New Jersey have jurisdiction over any dispute arising out of them, without prejudice to any right you have under the law of the place you live to bring proceedings there.
This is the one paragraph on the page that does not come from the code, and it would be dishonest not to say so. Every other statement here was written from something a reader can check — a line of a script, a constant in a file, a header on a request. A choice of law is not that kind of fact. It follows from where the publisher is formed and from a decision only the publisher can take, and it is stated above because the publisher supplied it, not because anything in this build implies it. If you are relying on it, write and ask rather than reading it off this page: clause 1 has the address.
Nothing in this clause affects anything else on the page: the copyright position in clause 2, the permissions in clause 4, the privacy description in clause 8 and the removal route in clause 11 all stand on their own and do not wait on it.
If a clause on this page is held to be unenforceable, the rest of the page continues to apply.
13 · Changes
Changes, and when this page last moved.
¶This page changes when the site changes. If a desk starts reading a new source, if a tool starts sending something it did not send before, or if what is stored in your browser changes, this page is updated in the same pass — that is the point of writing it from the code rather than from a template.
Changes take effect when they are published here. There is no mailing list for them and no banner: the date below is the record, and the clause addresses are stable, so a version you quoted can be compared against the one standing now.
This page last changed 19 August 2026. It was first published in August 2026.