← legal

IP Strategy — what protects a software product

Last updated: 2026-05-31 Audience: future-Sandon, advisors, investors who ask "what's your IP moat" or "shouldn't you patent / copyright / protect this?" Not legal advice. Working notes from founder-level research.

The advice "you need to get that copyrighted before someone steals the idea" comes up often, especially from people whose experience is in hardware, CPG, or other industries where patents and copyrights are load-bearing. It's well-intentioned and usually wrong in specifics, even when the underlying instinct is right.

What it actually protectsUseful for HiveJournal / Graphene?
CopyrightSpecific creative expression — the actual code, prose, audio recordings, designs, music. Automatic from the moment of creation. Registration adds litigation leverage (statutory damages, ability to sue) but isn't required for protection.Already protected automatically. Don't pay to "register" the platform's copyright unless we're planning to litigate — the protection exists regardless.
TrademarkBrand names, logos, product identifiers. Protects the name customers know us by. Registration with USPTO gives nationwide protection + presumption of validity.This is the one that matters for "name protection." Filing covered in TRADEMARK_FILING_KIT.md.
PatentNovel inventions / methods. For software, gutted by the Supreme Court's Alice v. CLS Bank ruling (2014). Most "AI orchestrates LLM and TTS" claims fail Alice's abstract-idea test.Not useful. Software patents are expensive ($10K+), slow (3+ years), narrow when granted, and rarely survive challenge for this kind of product.
Trade secretConfidential internal information — algorithms, customer lists, processes. Protected by NDAs and the act of keeping it secret.Limited. Most of what makes Graphene valuable (the AI orchestration pipeline, the content) is either public-facing or made of commodity components. Some internal algorithms (auto-show generation, character-voice autocasting) could be treated as trade secrets if we want — protected by NDA with contractors and by not open-sourcing those modules.
ContractsTerms of Service, employee/contractor IP assignment, NDAs.Important. A solid ToS prevents scraping/reverse-engineering claims; IP assignment makes sure code written by contractors actually belongs to the LLC.

The brutal truth about "stealing the idea"

Software ideas themselves are not legally protectable, no matter how specific or how good. Two examples that come up often:

  • Slack. Got cloned by Microsoft Teams within ~3 years of launch. Microsoft outspent Slack on distribution and now has more enterprise users. Slack didn't sue anyone — and didn't lose by being cloned; it lost by being outspent.
  • Zoom. Existed for 7 years before COVID. Google Meet, Microsoft Teams, BlueJeans, Webex, FaceTime all existed already. Zoom won by being the best product when video calling exploded, not by being first or by having patents.
  • Dropbox. Famous "this can be built with rsync" Hacker News thread. Got cloned by Google Drive, OneDrive, iCloud, Box. Still alive because of distribution and execution, not IP.

The "untouchable" software companies got cloned within months of launch. They won by moving fast, building distribution, and growing brand recognition — not by suing competitors.

What actually protects a software product

In rough order of leverage:

  1. Speed of execution. Whoever ships features users care about first wins the mental positioning. This is by far the biggest moat for an early-stage product.

  2. Distribution and brand. The audience knows us by name. Even if someone clones the product, the name "Graphene" + the existing listener base + the writers in write.cafe is a moat that takes years to recreate.

  3. Content library. Every audiobook published, every universe built (Turing Logs, Deep Cut), every cafe contest winner archived, every Bluesky author featured — that's content a copycat would have to recreate from zero. Months to years of work.

  4. Community. write.cafe writers who've built streaks and relationships, listeners who write letters that shape stories, Bluesky authors who got featured — these relationships are not transferable. A copycat could clone the software in a weekend; the community took years.

  5. Trademark on the name. Protects against a competitor trading on our recognition. The actually-protective legal filing — covered in TRADEMARK_FILING_KIT.md.

  6. Copyright on code and content. Automatic. Doesn't prevent someone from cloning the IDEA, but does prevent them from copying our actual code or audiobook recordings verbatim.

  7. Trade secret on the unique internal stuff if/when we have anything truly novel under the hood. Most of our pipeline uses commodity components (LLM API, ElevenLabs API, ffmpeg) — the value is in the integration and the creative direction, not in any single secret algorithm.

Why software patents barely work

The Supreme Court's Alice v. CLS Bank ruling (2014) created a two-step test that software patent claims have to survive:

  1. Is the claim directed to an abstract idea?
  2. If yes, does it have an "inventive concept" beyond just applying the abstract idea to a generic computer?

Most software claims fail step 2. "Take a journal entry → run it through an LLM → get an audiobook" is an abstract idea applied to generic computers (LLM API, TTS API). USPTO and courts both reject these routinely.

The patents that survive Alice tend to involve genuine technical novelty in the implementation — new algorithms, hardware integration, specific data structures that produce measurably better results. We don't have any of that. We have great integration of commodity AI components + great content + great brand — none of which is patentable.

Cost of trying anyway: $10–20K and 3+ years for a typical software patent application, with maybe a 30% chance of grant and a high chance of being invalidated if ever asserted in court. Not a useful investment for early-stage company resources.

What I'm actually doing (concrete actions)

ActionStatusDoc
Trademark filing — EMBERKILN via Point Seven Studio LLC, Classes 41 + 9, Section 1(b) ITUPlanned for filing 2026-05-31 (or shortly after)TRADEMARK_FILING_KIT.md
Terms of Service + Privacy Policy on graphene.fm and hivejournal.comNeeds draft before public launch(no doc yet — TODO)
Contractor IP assignment agreementNeeds draft before hiring contractors who write code or generate content(no doc yet — TODO)
Point Seven Studio LLC in good standing with its state of formationVerify before filing trademarkCheck state business portal
Copyright registration on key audiobooksOptional, low priority — only if we plan to litigate against a specific copycatNot currently planned

Conversation drafts

For a friend who says "you should get this copyrighted before someone steals it"

On the bigger "what if someone steals the idea" question — honest answer is that software ideas aren't legally protectable, no matter how good. Even patents barely work for software anymore (the law tightened a lot with Alice in 2014). What actually protects a software product:

  1. Trademark on the name → handling this week
  2. Copyright on the code and content → automatic from the moment we write it
  3. The content library — every audiobook, every universe, every writer in write.cafe is months of work a copycat would have to redo
  4. The brand and community — that's a moat that builds over years and can't be cloned

The companies people think of as untouchable — Slack, Zoom, Dropbox, Twitter — all got cloned within months of launch. They won by moving fast and building distribution, not by suing anyone.

So the protection IS real, but it lives in trademark + content + execution, not in a single "protect the idea" filing. The actual risk worth thinking about isn't theft — it's a better-funded competitor launching something similar and outspending us. Which is exactly why investor conversations matter.

For an investor who asks "what's your IP moat"

The defensibility doesn't come from legal IP — software ideas aren't really protectable and software patents are largely dead post-Alice. The actual moat is the content library, the brand, and the community. Three years from now we'll have hundreds of audiobooks in shared-lore universes, a write.cafe community of serious writers, and the EmberKiln brand on every channel. That stack is the moat. Trademark is filed defensively to prevent name confusion. Speed of execution and capital efficiency are what determine whether we get to that three-year position before someone else does.

Things I'd NOT do, for the record

  • Pay anyone $X00–$X000 to "register" copyrights on the platform before launch. Automatic protection exists; registration is only useful pre-litigation.
  • File a software patent. Will burn $10K+ and 3 years for protection that probably wouldn't survive an Alice challenge anyway.
  • Over-think the wordmark. "EMBERKILN" is a coined / fanciful word — inherently distinctive, the strongest category of mark — so it files as a bare standard-character wordmark with no §2(e)(1) descriptiveness risk. (This was a live concern under the old "Graphene Studio" name, where "graphene" being a known material forced a composite mark; the rename to a coined word made the problem disappear.)
  • Delay launch to "complete" IP work. The trademark application itself locks in the priority date the moment it's submitted — protection starts on filing day, not registration day. Launch on schedule with the application in flight.
  • Be defensive about cloners. Spend the worry on the actual risk (better-capitalized competitor outspending on marketing), not the imagined one (someone literally stealing the idea).

Sources

IP STRATEGY — Docs | HiveJournal