Meta wants your website in the bargain
Meta's revised commercial terms extend crawling permissions to websites associated with a business. The question is what businesses receive in return, and how freely they can refuse.
Field Notes
Thinking, in public.
Arguments, research and experiments on the future of the internet. A place to investigate technical systems, human behaviour and commercial consequences, and test the ideas connecting them.
What are you trading when you use a free LLM?
My view is that exchanging customer information and training rights for LLM access is a trade. The harder questions are whose consent covers it and whether that use is fair and reasonable.
Open models, unsettled economics
Open models are changing the price of intelligence and where control sits. The rapid shift in usage is another reason to question whether AI's economics have settled.
Ads won't be the backbone of the AI economy
AI will reach for the internet's familiar business model. I expect the more durable money to come from trusted information and transactions completed on our behalf.
Why static-first architecture matters more than schema in a post-search world
Schema is not the answer. Why static-first architecture (Astro, edge) wins for llm retrieval, search crawlers, and agentic browsing in a post-search world.
Web iv is the intelligent and agentic web
My thesis on the next era of the web: intelligent systems acting on our behalf, and the need to return value to the people who create it.
The economics of AI are yet to be determined
AI feels cheap. It feels cheap because investor capital is subsidising it. Four structural forces are converging that reframe the trajectory as an open question with significant downside.
What the RBA told us about marketing strategy for 2026
When AI told a consumer 'you have a good deal, don't switch', it wasn't providing information. It was giving advice. The inflation data raises a strategic question about need-led comparison.
AI discovery surface research: needs-driven economy insights
AI surfaces are not just answering questions; they're making recommendations, naming winners, providing decision-ready verdicts. Brands that don't appear are invisible at the point of decision.
ChatGPT advertising: what has changed and how to respond
ChatGPT has announced plans to test advertising in the U.S. This does not change current marketing in Australia, but it indicates where conversational AI is heading.
What the OAIC's privacy sweep signals for marketers
On the surface, the OAIC's first compliance sweep looks narrow. It is not. The regulator is shifting from reactive complaint handling to proactive market enforcement.
What 2025 taught us, and what 2026 demands
2025 was the year the old marketing playbook stopped working. 2026 won't be about incremental adaptation. It will be about reconstruction.
Why Australia's under-16 social media ban could backfire
The intent is sound. The mechanism is blunt. Once minors present as adults to evade the ban, the safety systems that protect them stop working.
Signal loss is accelerating, and waiting for legislation is the wrong play
Marketing effectiveness is being eroded by a structural decline in signal quality across the major platforms. The performance impact is happening before the regulatory impact.
Artificial testimonials and their impact on trust and compliance
AI-generated avatars presented as genuine customers erode trust and risk breaching Australian Consumer Law. A note on what this means for marketers and platforms.
TOON: a new way to structure data for language models
Token Oriented Object Notation is a compact, human-readable format for feeding structured data into LLMs. It's not a JSON replacement, but for the right shapes, it earns its place.
What dropping CTRs tell us about the state of search
Click-through rates are falling, and everyone's trying to figure out what it means. The signal is not decline. It is redistribution.
What AI whispering means in the context of agentic browsers
Hidden manipulation of AI agents through covert instructions. The user sees nothing unusual, but the agent reads an extra layer of meaning and acts on it.
Some learnings from MeasureCamp Sydney 2025
Analysts are no longer just interpreters of data. They're builders, product thinkers, and technologists. Four signals from a day at MeasureCamp.
Which API architecture powers the agentic web
GraphQL defines what is requested. gRPC defines how it is delivered. SOAP-based systems used additional standards to express terms. None yet define why. That missing layer is what comes next.
AIP vs MCP: context vs conduct
MCP is the internal wiring that makes the agent capable. AIP is the diplomatic code that makes the agent trustworthy. One enables thought. The other governs action.
The 12 surfaces of discovery
Search isn't declining; it's fragmenting. Mapping the twelve discovery surfaces where intent now meets information.
Reasoning, human vs artificial: structured systems and intelligence
Reasoning is less a human privilege and more a universal function of structured systems. Humans reason through awareness and intent; machines reason through scale and structure.
The end of web crawling (probably)
Crawling is built on the assumption that information is public and predictable. The web isn't that anymore. My hypothesis is that APIs will replace crawling as the dominant model for how data moves.
The agentic layer: how autonomous agents reshape the web by 2030
By 2030, the transactional layer of the web will be dominated by autonomous agents negotiating with brand APIs. Websites won't disappear, but they'll become backend infrastructure.
Thinking in layers: UX, AX, and the agentic web
Three definitions for concepts emerging in the near future. UX is where meaning meets mechanism. AX is where machine trust is built. AIP is how they talk.
The three layers of the agentic web
How the emerging agentic internet separates use, intelligence, and control. UX is human trust. AX is machine trust. AIP is institutional trust.
The Data Value Quadrant
A practical model for evaluating how data is used in marketing. Plot consumer value against data sensitivity. Decide what to proceed with, what to redesign, and what to retire.
The CUV Framework
Consent, Utility, Value. A simple, repeatable method for evaluating whether a data use case is ethical, valuable, and legally defensible.