Tech • AI • Robotics • Game
A developer just destroyed Adobe (and it concerns you)
7/10TL;DR
Cheap, capable AI is rapidly lowering the cost of cloning, reading and rewriting software, threatening business models built on opaque code, lock-in and maintenance rents.
KEY POINTS
Open-source clones challenge Adobe’s moat
Developer Brandon Thomas released seven open-source tools modeled on the Adobe suite, including alternatives to Photoshop and Premiere. The projects remain early-stage, but the significance lies in how a micro-team, aided by Claude, reproduced products that once seemed protected by decades of R&D, large engineering staffs and subscription lock-in. That weakens the case for premium pricing based mainly on proprietary implementation.
A broader pattern is emerging across software
Similar dynamics are appearing in games and enterprise code. Unofficial browser-based versions of Call of Duty: Black Ops and Halo show how decompilation and AI-assisted rebuilding can move software into new environments without the original publisher’s participation. In legacy IT, large COBOL codebases that long supported banking systems are becoming easier to read, modernize and migrate.
The old protection was compiled opacity
For decades, compilation did more than speed up software. It hid architecture and engineering choices while still delivering a usable product, allowing vendors to charge for licenses, anti-copy systems, integrations and support while customers could run programs but not meaningfully inspect them. That asymmetry underpinned ecosystems around Apple, Oracle, Nintendo, SAP, Sony and Adobe.
AI is climbing a “readability” ladder
The first vulnerable layer is the visible product itself: if the interface and behavior can be observed, an AI agent can infer requirements and rebuild a similar engine. The second is code too large or messy for humans to fully understand, where models can parse sprawling legacy systems. The third is raw binary without source code, where iterative rebuild-and-compare loops can gradually reconstruct equivalent code.
Game and device protections look weaker than before
AI-assisted reverse engineering is accelerating software porting and decompilation that once took enthusiast communities years. The process relies on automated loops that regenerate code, recompile it and compare the output against the original binary until it matches more closely. That threatens game exclusivity, embedded software secrecy and digital locks in devices distributed to customers.
Server-side systems are safer, but not untouchable
A fourth layer involves software users never receive directly, such as model APIs and recommendation engines. Here, imitation can happen through distillation, where large numbers of outputs are sampled to train a mimic system. Anthropic has described campaigns involving more than 16 million interactions from 24,000 fake accounts, highlighting how cloud delivery reduces but does not eliminate exposure.
Hardware and tacit knowledge remain stronger defenses
Two higher layers are harder to copy. Physical chips can be reverse engineered, but reproducing them still requires very expensive manufacturing capacity. Tacit knowledge, such as undocumented operational know-how, customer context and unwritten rules, remains the most resilient because it was never fully encoded in software in the first place.
The likely winners are firms with hard-to-copy assets
As readable code loses scarcity, defensibility shifts toward four areas: physical infrastructure, time-based trust such as compliance and reputation, legal protection such as brands and ecosystems, and execution speed. This is driving a “flight to the cloud,” more emphasis on hardware-enforced controls, and greater reliance on trademarks and distribution gatekeepers rather than secrecy alone.
Several business models are especially exposed
One is software-locked hardware, where features already present in a car, printer or machine are paywalled by code. Another is the maintenance toll, where vendors profit from high exit costs more than software quality, as seen in infrastructure and no-code platforms such as Bubble, Power Apps, Mendix and OutSystems. A third is IT services built around scarce expertise in unreadable systems, from Accenture and Capgemini to legacy support specialists.
Finance may also absorb part of the shock
Investors and lenders have poured roughly $500 billion into software businesses on the assumption that customers were effectively trapped. If AI lowers migration costs or even makes credible threats to leave, customers gain leverage in renewal negotiations, margins come under pressure and the value of revenue streams tied to lock-in may have to be reassessed.
The implications extend beyond software
Industrial software can encode years of tuning knowledge, operational shortcuts and engineering trade secrets. As AI makes that code easier to inspect, competitors may extract valuable know-how from shipped products, from machinery to chip software stacks such as Nvidia CUDA. The same logic is beginning to appear in biology, where the cost of reading DNA has fallen from roughly $3 billion for the first human genome to only a few hundred dollars today.
CONCLUSION
The central shift is from protecting what is hard to understand to protecting what is hard to physically reproduce. Companies whose margins depend on opaque code, captive formats or costly exits face mounting pressure as AI turns readability into a commodity.
Ask a question