You can’t predict the future. You can build the fitness to respond to it. #MindCandy is fuel for strategic conversation about what’s next. Forward this to your leadership team and DM me for a conversation.
Would you buy AI from China
Two signals:
- What would have to be true for you to cut your coding AI bill by 90%? Revolut says it moved about 2,000 engineers from Claude Code and Codex to GLM 5.3, a Chinese model it hosts itself. The 90% is the reported saving on coding AI. The figure I’d want next is the cost of getting a working change into production, including the engineer’s time and any rework. Cybernews
- Western startups are already adapting Chinese open-weight models for businesses concerned about censorship and security. That gives buyers three separate choices. Who built the model? Who adapted it? Who runs it? Each answer matters when something goes wrong or a supplier raises its prices. Rest of World
Pick one recurring task before your next renewal. Give your current tool and a Chinese open-weight alternative the same ten examples, using data you can safely share. Have the people who do the work judge the outputs blind. Count hosting, review time, and rework, then compare the cost per result your team would actually use. Take that number into the renewal meeting.
The other pompts:
Why you’re probably spending too much on software (and AI)
European small and medium-sized enterprises are overspending on software and AI by an impressive €10bn per year.
https://www.raconteur.net/technology/why-youre-probably-spending-too-much-on-software-and-ai
Turning dark data into an AI advantage
Industry estimates suggest as much as 80 per cent of enterprise data remains ‘dark’ – uncatalogued and unused information that is scattered across legacy systems or buried in scanned documents and email archives. Meanwhile, just under one per cent of it has made it into AI models.
https://www.technologyrecord.com/article/turning-dark-data-into-an-ai-advantage
The long bill for cheap software
Roughly 44 per cent of AI code-generation tasks introduced a security flaw.
https://www.businesstimes.com.sg/startups-tech/technology/long-bill-cheap-software
How AI and Edge Computing Are Turning IoT Data Into Real-Time Business Decisions
Real-time processing has become such an important part of IoT. A 2025 survey of more than 600 enterprises across 10 countries, found that 82 percent were already using or planning to implement real-time data processing for IoT.
https://www.iotforall.com/ai-edge-iot-data-real-time-decisions
Borderless data architectures: Empowering AI agents where your data lives
AI agents need intelligent access to data wherever it resides. That makes it increasingly important to bring modern AI and data capabilities across cloud, edge and private data centers rather than requiring every workload and dataset to move toward a single destination.
New signs of rising risks
Cyber chaos, AI bubble bursts, AI hits true recursive self-improvement, war expands, oil spikes and politics.
https://www.axios.com/2026/10/25/ai-cybersecurity-politics-risks-2028
More than an AI copilot: Navigating the 5 stages to autonomous data engineering
The Autonomous Data Engineering maturity curve
How CIOs Can Govern the Rise of Citizen-Built Applications
Citizen development introduces unexpected security risks. A data breach from an unapproved application can force IT teams into firefighting mode as they scramble to identify the source and contain the damage. Routine security scans may uncover critical vulnerabilities in applications that the business and information security didn’t know existed.
https://www.fticonsulting.com/insights/articles/how-cio-can-govern-rise-citizen-built-applications
The Silent Failure In Enterprise Data: When Systems Stay Up But Meaning Breaks
A successful pipeline creates a governance failure when a definition changes without downstream users knowing about it.
Who wins when AI does the shopping? New evidence on agentic commerce
Change Management For Agentic Data Modernization
Agentic AI raises a larger question: Are we ready for software to make decisions for us?
Anthropic warns that it’s very close to RSI capability, aka self-building AI
RSI is the point at which AI grows so good at training itself that it spawns a dramatic acceleration of AI capabilities. To the AI doomers, RSI represents the cataclysm from which there is no return, in which AI builds itself into something powerful enough that it slips beyond human control and may be indifferent to the damage it causes while achieving its goals.
