Mighty Bear Games exits games as SecretSauce Labs, selling AI brand-consistency tooling to enterprises
Summary
Wearemighty, formerly Mighty Bear Games, rebrands for the second time in under a year, becoming SecretSauce Labs and exiting game development entirely for enterprise AI content tooling.
The pivot traces to a late-2022 production bottleneck: the studio needed roughly 25 million distinct, on-brand playable avatars, a volume early generative tools could not produce consistently.
Three years of internal R&D turned that consistency problem into a standalone product built around checking AI output against brand material before a human sees it.
"Ultimately we're here to sell outcomes, not resell inference"
01
A production crisis becomes the founding thesis
In late 2022, Mighty Bear Games needed around 25 million distinct, on-brand playable avatars for a game in production, a volume manual creation could not support.
Early generative tools produced fast, creative output but no continuity between requests, sometimes altering a character's face or outfit or adding an extra finger on repeat generations.
"We would ask for the same character twice and get a different face or outfit, or sometimes an extra finger"
Davis had experimented with machine learning since 2012, but the avatar volume problem forced the team to treat consistency as an engineering problem rather than a creative one.
02
The claimed defensibility rests on catching brand drift before humans see it
Davis treats storage and access to brand assets, logos, style guides, and past creative, as commoditized, since many existing tools already support uploading that material.
The layer he calls mostly unsolved is checking every new AI generation against that brand material and flagging a miss before a human reviews it.
Of the three problems SecretSauce Labs set out to solve, scale, brand consistency, and cost, Davis calls brand drift the hardest, requiring three years of production work with creative experts from major gaming companies.
"What's mostly unsolved is checking every new generation against that material and catching a miss before a human ever sees it"
03
Business model claims outpace disclosed unit economics
Davis frames the pivot as selling outcomes rather than reselling model inference, positioning SecretSauce Labs as a margin layer above commodity generative-AI providers.
He cites a productivity example: work once requiring an art team of six people over three months can now be completed by one person in a day, but only "in the right circumstances."
Davis himself qualifies that claim: generating ten assets and being able to use only two of them is not actually faster once editing and brand fixes are counted, and can cost more than doing the work properly the first time.
No client count, contract value, pricing model, or revenue figure for SecretSauce Labs appears in the interview, leaving the outcome-based economics unverified against Davis's own caveats.
"Most of the 'AI saves time' story ignores that generating 10 assets and being able to use two of them isn't actually faster"
04
Internal adoption resistance and rapid rebranding signal execution risk
Davis acknowledges that inside studios adopting the tooling, teams sometimes slow-walk implementation or sandbag results out of fear that working tools make them redundant.
He states the effort is about building team capability while teams keep their own workflows and standards, and does not involve cutting headcount.
That adoption resistance is the same organizational hurdle SecretSauce Labs must now sell against inside every enterprise brand team it targets, beyond the game studios where the problem originated.
The company has changed identity twice in roughly a year, from Mighty Bear Games to wearemighty in January 2026 and now to SecretSauce Labs, a compressed rebranding cycle that suggests the strategic positioning is still being finalized even as it is pitched to customers.
"People are scared that if the tools work too well, they become redundant"
What this means
For Investors & VC: SecretSauce Labs discloses no clients, contract values, pricing or revenue after two rebrands in a year, so outcome-based margin claims require audited unit economics before any valuation.
For Publishers & Developers: When evaluating AI art tooling, measure usable-asset yield after editing and brand fixes, and plan for internal teams slow-walking adoption over redundancy fears.
For Service Vendors: Brand-asset storage and upload is now commoditized; the defensible layer is automated checking of each AI generation against brand material before human review.
Events
Companies
Games
—
Locations
—
COMPILED BY GAMES ATLAS EDITORIAL
RELATED ARTICLES
EXPLORE MORE
Comments (0)
No comments yet. Be the first to share your thoughts!