How the Latest AI Leap Is Disrupting Business Models

AI systems that once seemed to improve only incrementally are now being used to decompile software and rebuild it in new forms, putting pressure on assumptions about how software companies earn money. Users have applied Astra and Opus V5 to games, while Claude Opus enabled Brandon Thomas to reconstruct Photoshop through a clean-room rebuild. The business implications extend beyond cheaper development: if individuals can reproduce useful software or generate tailored experiences, established companies may need to compete on more than access to code.

The Financial Times reported that Anthropic told investors it generated $16.8 billion in Q3 revenue, up 45% from Q2; that was quarterly revenue, not an annualized run rate, and likely preceded revenue shares with Amazon and other compute partners. Anthropic also reported adjusted profitability, positive free cash flow, 85% adjusted gross margins, and 11% adjusted operating margins, but its adjusted figures have typically excluded stock compensation, training costs, and revenue shares. Reuters reported that Anthropic expects $110 billion in annualized revenue by year-end, versus about $70 billion based on Q3, implying 57% growth in Q4. Ticker Trends showed growth flat since late July, while Andrew Curran suggested the forecast might reflect confidence in an upcoming model release. Anthropic was reportedly planning an IPO at a $2 trillion valuation in the coming months, potentially using disclosures limited to Q1 and Q2.

Thomas said he had “one-shotted Photoshop” and later offered AI-generated versions of Adobe’s suite, making a formerly costly, time-consuming reconstruction more practical for an individual. The recreations still had bugs and lacked features compared with Adobe’s software, priced at $100 a month; Adobe shares opened 3% lower after the story went viral, then partly recovered. Chamath Palihapitiya argued that headless software accessed through agent tools and MCP could make software IP effectively worthless, and said his AI SaaS business had stopped negotiating to retain IP in client systems. That may overstate the threat: piracy has not prevented Adobe from continuing its business, SaaS providers also sell support and maintenance, and earlier attempts to replace custom CRM and accounting systems have not broadly displaced contracts, with many users returning to Salesforce.

Gaming faces both weak market conditions and the prospect of cheaper, more personalized production: console sales are at a 13-year low, tens of thousands of studio jobs have been cut, and many new games have sold poorly. Bloomberg Intelligence said streaming and AI could eventually make consoles one of several relatively uneconomic ways to access games, though consoles may persist for enthusiasts and collectors. A base PS5 costs $650, up from $500 at launch, while the Steam Machine is cited at $1,000; AI-accelerated jailbreak efforts have also enabled PS5s to play games from other systems. Andrew Curran said recent models can produce stable, playable game mashups in a day and bring one-person studios closer. He predicted that personalized, AAA-level games could undermine demand for studio-made titles if generation scales; George Cal forecast solo developers could ship high-quality games within one to two years, while Elon Musk said “next year.”

Companies are already changing how they build. Epic Games banned handwritten code and moved toward AI-first production, including entirely AI-generated art workflows for a delayed Disney shooter featuring Marvel and Star Wars characters. Tim Sweeney said the goal was to help teams make more and better work, not simply the same work with fewer people; Andre Apansik criticized the move as failing to “read the room,” while Sweeney said Epic had to build for current and future conditions. The push has had rough edges: the new launcher was described as “vibe coded,” and gamers reported visual glitches in a fast-launched release. Xbox, meanwhile, created Xbox Labs, a small skunkworks for emerging ideas, including AI experiences for players and developers. Christian Catalini said Xbox could respond to decompiled games through legal action or by embracing change through a lab.

AI is also shaping bets on existing businesses. Ithaca Holdings completed a take-private deal for Bumble after assessing hundreds of companies for value as AI advances. Ravi Gupta said Ithaca believes abundant intelligence could make meaningful relationships more precious and frontier technology could help people form them; Bumble’s valuation reached $14 billion after its 2021 IPO before its stock price fell 95% the following year. Bending Spoons is pursuing a different thesis: AI and staff aggregation could lower maintenance costs at legacy brands that still have revenue and viable products. Berkshire Hathaway agreed to invest $330 million after putting in $600 million during Bending Spoons’ July IPO, bringing its stake to 4%; the stock rose 40% on Wednesday after the news.

David’s Bridal is pivoting from selling wedding dresses toward an AI-powered planning service it calls “aisle-to-algorithm.” CEO Kelly Cook described the asset-light strategy as a major bet and a 180-degree turn; the planner connects brides to a vendor portal and draws on information from appointments, the website, and other services. Elena Vilk said agentic coding tools and a lean team reduced work that once would have required more than 100 people to work done by tens. The company filed for bankruptcy in 2023 after a weak pandemic recovery and is seeking a larger share of the wedding industry amid a long-term decline in marriage, although AI itself has not disrupted the wedding-dress business.