Amber

Amber is forever media hosting. Pay once. Hosted for life.

The phrase in my head is freezing media in amber. Not because the internet should become a museum, but because links should not rot just because a platform changed its pricing page.

I built a site for my dad. It should not stop existing because a credit card expires in 2031. A photo album does not bill your estate monthly. Your web things deserve the same stability you would want for a box of letters, and somehow we all agreed to rent them instead.

The idea

The secret the platforms do not advertise is that most media does not cost much to keep alive. The expensive part is not a human lifetime of storage. The expensive part is the business model wrapped around it.

The way I said it into my voice recorder in March:

“the average YouTube video costs Google nothing over a lifetime… I think what I am going to be doing is fifteen dollars per hour of video for lifetime hosting.”

That $15 per hour of video is my working number for lifetime hosting. Not a live offer. A planned price that keeps me honest while I test the endowment math.

The bet is simple: if the real cost of storing and serving ordinary media over a human lifetime is a couple of dollars total, then the humane product is not another subscription. It is an endowment. Pay once. Put enough aside to keep the file alive. Make the policy inspectable. Do not punish people every month for wanting their father’s old website, their short film, their podcast archive, or their family videos to keep existing.

People think they are buying hosting. What they are actually buying is confidence that the thing they made will still be there in twenty years. That confidence is the product. The servers are an implementation detail.

Amber sits under Sherwood because this is exactly the kind of infrastructure I want owned by people with long time horizons.

What exists today

• A working prototype exists with a substantial codebase.
• Static site hosting works, including custom-domain routing.
• ZIP deploys work as the basic publishing path.
• There is a magic importer that crawls a live site into a deployable copy.
• The Amber Player supports multi-angle video.
• The player supports stem mixing, so a piece of media can carry separate audio parts.
• The player has automatic fallback when a viral-traffic circuit breaker trips, so one viral video cannot bankrupt the endowment.
• Podcast hosting exists with RSS 2.0 and Podcasting 2.0 support.
• Existing podcast feeds can be imported in one call.
• Channels now cover podcast, video, and music feeds, with RSS, JSON Feed, WebSub pings, and publish-from-asset support.
• The media dashboard can upload, browse, play, unpublish, reactivate, withdraw, and publish hosted assets into channels.
• There is an agent-readable API surface: /api, /api/openapi.json, /llms.txt, structured errors with hints, cursor pagination, idempotency keys, and a small CLI for deploys, uploads, imports, quotes, and webhook setup.
• Lifecycle rules are wired through the product: inactive sites, feeds, assets, manifests, and enclosures return gone instead of quietly serving withdrawn work.
• The backend has the bones of durable operations: entitlement records, webhook subscriptions and deliveries, worker leases, account deletion that does not erase purchases, and export manifests.
• Policy snapshots are versioned, so the forever promise can be inspected later instead of remembered vaguely.
• The first patient is my dad’s old Flash site about walking the Finger Lakes. I refuse to let that website die.

That first patient is important to me. Not because it is commercially clever. Because it is the correct emotional test. If Amber cannot save a weird old family website, it is not doing the job I need it to do.

What it isn’t yet

Amber is not publicly launched. The product has been live-verified in development with real media, but there is no public domain I am asking people to use, no public checkout flow, no customer support process, and no legal policy I would ask a stranger to rely on yet.

The endowment math is also not finished. Storage costs are easy to wave around and hard to promise against forever. Bandwidth spikes matter. Abuse matters. Copyright fights matter. Maintenance matters. Human attention matters.

The lifetime promise has to be boring enough to survive me having a bad week. It is not there yet.

Where it’s going

The plan is to make Amber the place for media that should keep its URL. Films. Podcasts. Old websites. Family archives. One-off projects. Things that do not need a growth team, just a durable address.

The planned pricing I keep testing against is $15 per hour of video, lifetime. Again: my working number, not a live offer. Static sites and audio have their own math, but the principle is the same. Charge once. Reserve enough. Keep serving.

The failover design is part of the philosophy, not a compromise of it. Platforms make fine pressure valves and terrible landlords. If a video goes viral, Amber can route the flood to YouTube while the original stays home at its own address. Distribution is not home.

I also want DMCA resilience for legitimate fair use. If a fair-use file is wrongly removed, the creator should be able to restore it at the same URL so the citation does not break. That is not a promise I can make today. It is a design requirement.

Bring-your-own-storage is part of the plan too. Amber should not require trust in one bucket, one provider, or one person. The forever promise has to be checkable.

Like everything I make, the recipe will be free: how to move a small site off a platform, how to host video without handing YouTube the master copy, how the endowment math actually works. I’m writing those up as I go, and they land on the Pantry. Paying Amber is for people who would rather not run their own forever.

The sibling project is Elsewhere, because owning your words and hosting your media forever are really the same rebellion from two different doors.