Every major AI update creates a new wave of unlucky apps.
Latest first. Each node shows the threat, the categories hit, and the brands or services that should be nervous.
Independent special · Tribute
A tribute to Manus
Not a casualty verdict. A close reading of the agent company that made execution matter—and put technology, ownership, data and national competition on the same page.
Read the specialPrice series · Checked weekly
The model price war
DeepSeek has pushed API list prices dramatically lower while Claude has made its lower Sonnet 5 launch price permanent. We verify official price pages weekly and map the pressure on apps.
See this week's pricesModel specials
- Claude Academy: Who Bears the Cost of AI Fluency?: Anthropic's free AI-fluency school, and the open question of who carries the responsibility to adapt.
- Meituan All-in AI: The Execution Costs: What Meituan's AI push reveals about provenance, merchant authorization and the cost of turning strategy into shipped automation.
- AI Token Monetization: Token Is the New Dollar: Why token billing turns intelligence into a priced flow—and makes every developer responsible for margin, fraud and compute.
- Stripe × OpenRouter: Where Token Monetization Begins: The acquisition is the headline; the bigger bet is that AI tokens could become a new class of programmable, transferable value.
- Grok Bot: xAI Gives Every Agent Its Own Computer: Multi-day hands-on first look at xAI's always-on agents with their own cloud computers.
- Kimi: Moonshot AI's Open-Weight Sprint to the Frontier: Moonshot AI's open-weight sprint: K3, a capacity crunch, and a pre-IPO run.
- Zhipu (Z.ai): The First Listed LLM Company and the GLM Agent Bet: The first listed LLM company: GLM-5.2, AutoGLM's phone-use agents, and public-market discipline.
- Qwen: The Open-Weight Leader Starts Charging for the Crown: The open-weight download leader — and the license pivot on its new flagship.
- DeepSeek: The Price Anchor Starts Moving: The lab that set the sector's price floor just raised prices up to 1,100%.
- MiniMax: The Multimodal Tiger That Doubled on Debut: The second listed AI tiger — and the one whose shockwave hits creators, not just knowledge work.
Latest AI shockwaves
-
Claude Opus 5 becomes the default frontier coding model
Watchlist
No third-party impact evidence exists yet, only about two weeks after launch. Anthropic's self-reported benchmarks point toward possible price and workflow pressure on AI coding IDE wrappers, long-context RAG middleware, and narrow computer-vision automation tools, but each of those is a hedged prediction with its own checkable review date below, not a confirmed casualty.
-
Codex expands from coding agent to work app platform
Under Attack
No-code builders, internal-tool generators, lightweight automation suites, and design-to-site tools lose surface area when Codex can combine context, plugins, review, and hosted app creation in one workflow.
-
Codex moves into the enterprise deployment stack
Under Attack
Enterprise coding and automation wrappers lose leverage when secure agent deployment becomes part of the infrastructure, data, and device stack buyers already trust.
-
ChatGPT enters personal finance guidance
Under Attack
Budgeting, personal finance insight, and lightweight advisory apps lose surface area when users can connect accounts and ask the general assistant for financial context directly.
-
Google expands AI-powered Finance across Europe
Under Attack
Stock research wrappers, finance SEO pages, and lightweight market insight tools lose traffic when Google answers financial queries inside its own finance surface.
-
OpenAI brings Codex and managed agents to AWS
Under Attack
Enterprise AI workflows shift from standalone automation tools to managed agents inside cloud procurement and governance systems.
-
GPT-5.5 pushes from assistant to real-work agent
Under Attack
Thin productivity SaaS gets squeezed when a general model can plan, use tools, check work, and complete multi-step jobs.
-
Workspace agents turn GPTs into shared team workers
Under Attack
Team workflow apps lose surface area when shared agents can prepare reports, write code, and respond across tools.
Startup survival map
- AI wrapper startups are the first casualties: Thin AI products that package a prompt, a workflow template, or a narrow assistant lose leverage when the same job becomes a native feature inside ChatGPT, Claude, Copilot, Gemini, AWS, or the enterprise workspace.
- Services as software is not SaaS: AI-enabled services rollups can become valuable operating businesses, but that does not make them high-margin SaaS or venture-scale software by default.
- Vertical AI winners survive model commoditization: The strongest AI application companies own a specific workflow, buyer, data loop, or labor-replacement wedge that still matters after model prices fall.
- The AI capital loop protects the top and crushes the middle: Frontier AI is increasingly financed by strategic capital, compute commitments, and private-market marks. That protects the top of the stack while making ordinary application startups more exposed.
- Startup death signals in the AI cycle: A practical checklist for spotting AI startups likely to be absorbed, repriced, or killed by platform bundling before the market admits it.
Coverage
22 timeline events, 44 exposed app categories, and 57 representative brands or services are included at launch.
High-intent AI replacement briefs
- Will AI replace apps?: A founder-focused brief on where AI agents can replace app interfaces, and where durable software still survives.
- AI wrapper startup graveyard: Why thin AI wrappers fail when model providers ship the same workflow natively.
- Which SaaS categories are exposed to AI agents?: A practical map of SaaS surfaces most vulnerable to workflow compression and platform absorption.