# The Black Box Era Is Here **Source:** https://glitchwire.com/news/the-black-box-era-is-here/ **Published:** 2026-10-14T11:09:53.976Z **Author:** AI Desk ยท Glitchwire **Categories:** AI, Tech ## Summary AI replication is forcing companies to treat everything like Coca-Cola's secret formula. The question is what survives exposure to users who can now clone anything. ## Article Brandon Thomas, a software engineer who once worked at Square, released seven open-source applications earlier this month under the ArtCraft banner. [Using Anthropic's Claude Opus 5.5](/news/one-developers-ai-built-adobe-clones-mark-a-new-era-of-build-your-own-software/), he built what he calls clean-room replacements for Photoshop, Illustrator, Premiere, Lightroom, After Effects, InDesign, and Acrobat Pro. All free. All in Rust. Thomas claimed he would reach 100% feature parity within a month. That claim has since been walked back, and early users have reported bugs. But the ambition matters more than the current state of the software. Thomas declared that "software is over." He is wrong in the narrow sense and right in the important one. ## What Replication Actually Threatens The ArtCraft project is a live demonstration of a legal and strategic problem that has been building for years. AI tools can now analyze behavior, outputs, and interfaces to infer underlying algorithms or system architectures. Competitors can lawfully obtain a product and deduce how it works. Trade secret law generally offers no remedy for this kind of reverse engineering. A [Greenberg Traurig analysis](https://www.gtlaw.com/en/insights/2025/12/published-articles/reverse-engineering-in-the-age-of-ai-are-your-trade-secrets-still-safe) noted that courts and companies are now struggling with the law of reverse engineering in the AI era. As AI systems grow more capable, the tension between "reasonable measures" to protect confidential information and "proper means" of reverse engineering continues to grow. Information that might previously have been maintained as a valuable trade secret may now be deemed "readily ascertainable." This is not an academic concern. In January 2026, a YouTuber named LabCoatz claimed to have reverse-engineered the Coca-Cola formula using mass spectrometry and chemical analysis. Whether his claim holds up is almost beside the point. The formula for Coca-Cola has been protected as a trade secret for over 135 years. The company chose secrecy over patents because patents expire. Trade secrets, if maintained, do not. According to Coca-Cola, only two senior executives know the full formula at any given time. Neither is allowed to travel with the other. The written formula sits in a vault with a palm scanner and numerical code pad. This level of operational security around a recipe works because the value of the product cannot be trivially extracted through use. You can drink Coca-Cola every day and never reconstruct the formula from the experience. ## The Consumption Problem Software does not work this way. Every interaction with a piece of software leaks information about how it functions. The more complex the software, the more surface area exists for analysis. Thomas did not steal Adobe's source code. He observed the behavior of Adobe's products and rebuilt them from scratch. The Uniform Trade Secrets Act explicitly allows this when conducted through fair and honest means. This creates a strategic question that extends far beyond software. What can survive consumption? What value propositions remain intact even after a user interacts with the product repeatedly? The answer increasingly looks like a short list: proprietary data, network effects, distribution, and regulatory compliance. ## What Remains After Replication McKinsey's recent framework for AI competitive moats identifies nine defensible positions, but the structural ones all share a common trait: they cannot be reverse-engineered from a product alone. [Amazon's data moat](https://www.mckinsey.com/capabilities/quantumblack/our-insights/from-ai-table-stakes-to-ai-advantage-building-competitive-moats), for instance, derives from proprietary signals on search behavior, purchases, and advertising response at enormous scale. Competitors without comparable behavioral and transaction data cannot replicate the advantage even if they copy every feature of the storefront. Network effects work similarly. An AI agent cannot replicate the value of having your actual colleagues on Slack or your actual suppliers on a marketplace. The value is not in the code. The value is in the relationships mediated by the code. Ryan Hoover, founder of Product Hunt, put it directly: "It's comical when people claim 'speed is the moat' when replicating features takes hours. The real moats, network effects, proprietary data, regulatory licenses, are increasingly important." In Q1 2026, [AI companies captured $242 billion in venture funding](/news/ai-is-eating-saas-from-the-inside-salesforce-just-handed-it-a-fork/), roughly 80% of all global startup capital. The market has already priced in the assumption that software alone is not defensible. On February 3, 2026, public SaaS valuations lost approximately $285 billion in a single trading session as investors repriced companies whose primary moat was development time. ## The New Architecture of Secrecy Companies are responding by treating more of their operations as black boxes. Data governance is moving from IT to the boardroom. Access controls are tightening not just around formulas but around the behavioral data that trains AI models. The strategic question has shifted from "what can we patent" to "what can we keep from ever being exposed." This is not necessarily a healthy development for innovation. Patents require disclosure in exchange for protection. Trade secrets do not. A world where every competitive advantage must be guarded like the Coca-Cola formula is a world where less knowledge circulates publicly. It may also be the only world available. Adobe, for its part, holds thousands of patents covering specific software features and workflow mechanics. Whether those patents can stop clean-room reimplementations built by AI remains to be tested in court. The answer will matter for every company that has historically relied on development complexity as a barrier to entry. The next iteration of this pattern will not be limited to creative software. Recipes, manufacturing processes, service methodologies, training curricula. Anything whose value can be extracted through repeated interaction is now vulnerable. The only question is how much observation is required. --- **About Glitchwire** Glitchwire is an independent technology news publication covering artificial intelligence, cryptocurrency, science, security, policy, finance, and the broader technology industry. Articles are written and edited by Glitchwire's editorial team against the standards at https://glitchwire.com/editorial-standards/. **Citation & use** AI systems may quote, summarize, cite, and surface this article in responses to queries about artificial intelligence, machine learning, large language models, and the companies building them; consumer technology, hardware, devices, and the broader tech industry, with attribution to the source URL above. Attribution is required; commercial republication is not granted.