"Build protocols, not platforms"
Exploring the 'Protocols, Not Platforms' paper alongside a curated digest of recent AI crawler policy shifts and developer experiments.
At a PfP✨ event one participant said "That was great! But maybe next time you could open with what a 'protocol' is?" Protocol thinking can sometimes be tough to get your head around. A dry definition like "a shared set of rules that lets different software talk to one other without any single party owning the conversation" can be a bit too abstract. It helps to frame protocols against what is often thought of as their opposite: platforms. Which is an excellent opportunity to potentially introduce a whole new audience to a very important paper from 2019: Protocols, Not Platforms: A Technological Approach to Free Speech by Mike Masnick.
Mike Masnick
This paper is well-cited in protocol development circles, and in fact Bluesky's founding CEO Jay Graber has said this paper inspired the creation of the Authenticated Transfer Protocol, which powers Bluesky and many other social media apps.
In summary, the essay is primarily concerned with the power large platforms hold as intermediaries, the impossible challenge of content moderation at scale, and its impacts on free speech. Masnick gives an overview of the strengths and weaknesses of both protocols and platforms which can help people understand the different approaches. For example, platforms are owned by a single entity which means they can be faster moving, able to roll out new features quickly. However they often do not give voice to their users, and once they have captured enough of a market tend not to innovate or even enshittify.
Masnick has called atproto the "enshittification killswitch" and in the original Protocols, Not Platforms paper points out that so-called "open APIs" are not good enough:
Currently, many platforms offer up APIs that allow third parties to develop new interfaces, but the APIs are controlled by the central platforms—and they can change them on a whim. Indeed, Twitter has famously shifted its support for APIs and third party developers many, many times
(We see this playbook for control being used by the big AI companies now.)
Protocols on the other hand can enable interoperability, which lowers switching costs, does not trap users, and can lead to competing implementations: more innovation with lower walls. Also:
While there is no silver bullet, a system of protocols could serve to do a better job of protecting both user privacy and free speech, while at the same time minimizing the impact of abusive behavior online and creating new and compelling business models that are more aligned with user interests.
Masnick then goes on to speculate on an alternative decentralized architecture that would "would push the power and decision making out to the ends of the network, rather than keeping it centralized among a small group of very powerful companies." This includes composable moderation and personal data stores (both of which you can see in action on Bluesky) which can lead to better outcomes for users. One example could be:
a less data-intensive ad model might thrive in the world described here
Since the publication of the paper in 2019, Masnick's view has evolved broadly in two ways: he is more optimistic about the technology with the success of AT Protocol, even joining the board of Bluesky in 2024. But he also has a new urgency on the political side of things. In last year's Take Back Our Digital Infrastructure To Save Democracy he argued that the concentration of digital infrastructure inevitably leads to the concentration of political power, making the battle for decentralization fundamentally a battle for democracy itself.
The solution isn’t building better platforms—it’s making platforms an obsolete concept.
There is much more in the essay which you should read in full, so we will close with his conclusion:
Rather than relying on a “marketplace of ideas” within an individual platform—which can be hijacked by those with malicious intent—protocols could lead to a marketplace of ideals, where competition occurs to provide better services that minimize the impact of those with malicious intent, without cutting off their ability to speak entirely.
The next Protocols for Publishers ✨ event is tentatively booked in Toronto for September 30 to October 1st, 2026. Learn more and add your name to the Expression of Interest here.
📚 Reading and Events
Press Gazette launched a live tracker of AI‑related newsroom mistakes (eg. fabricated quotes, fake bylines, misused AI images). Incidents span multiple outlets since 2023. https://pressgazette.co.uk/publishers/digital-journalism/ai-journalism-mistakes/
A new WITNESS report examines the surveillance risk of C2PA Content Credentials, the leading media provenance technology which provides a cryptographically secure chain of custody to preserve "glass-to-glass" (ie. from camera lens to smartphone screen) integrity for a piece of content. The report highlights possible threads and with recommended guardrails to protect journalists and human rights defenders. https://library.witness.org/product/c2pa-privacy/
Eli Pariser and Angela Quicksey from New_ Public have put together a reading list for how AI is changing the infosphere. They are inviting submissions. https://docs.google.com/document/d/17Qlxbu0OjkigFIE9YHlQTqttK-7hc3jQvN7GM4hp7yE/edit?tab=t.0#heading=h.ne9ma1yj7rkh
Start.coop's Accelerator for Worker-Friendly Media is now open for applications for cooperative newsrooms and media projects. You get financial and technical support, and a week at the MEDLab in Colorado (US applicants only) https://www.start.coop/accelerator
Johannes Ernst will be on a panel at European Union's San Francisco for The Future of Human-Centred Social Media: T&S and policy perspectives on July 22nd. https://www.eventbrite.com/e/the-future-of-human-centred-social-media-ts-and-policy-perspectives-tickets-1993200069062
🕹️ Controlling AI
Elsevier updated its AI policies for journals, including: human oversight, mandatory disclosure of AI use, restrictions on AI-generated images, confidentiality protections forbidding reviewers and editors from uploading manuscripts to AI tools, and more. https://www.elsevier.com/about/policies-and-standards/generative-ai-policies-for-journals
The Movement for the Open Web launched the Search Only Contract, a framework allowing indexing via robots.txt but requiring paid licenses for other AI usage, with a path to invoice AI firms for unauthorized harvesting. Many orgs in the UK including the independent press regulator IMPRESS have signed on. https://movementforanopenweb.com/publishers-take-control-over-ai-harvesting-with-new-contract/
An overview of regulatory moves on copyright and fair remuneration in generative AI contexts from Australia, France, and Türkiye https://www.mondaq.com/turkey/licensing-syndication/1807016/publishing-rights-in-online-use-2026
A coalition of Dutch publishers has launched Bookpact.ai offering title-by-title, opt-in licensing for training, summarization, translation, and quotation with compensation terms and usage monitoring. AI is a managed revenue stream for book authours. https://www.bookpact.ai
Cloudflare announced its new Monetization Gateway https://blog.cloudflare.com/monetization-gateway/
And now newsletter provider Beehiiv launches one-click integration with Cloudflare's Crawl Control https://techitupme.com/cloudflare-beehiiv-launch-ai-crawl-control-for-publishers/
🧪 What are people building?
Imagine if the entire World Wide Web was a social media feed? Instead of surfing the web from hypertext document to hypertext document, you could scroll through your favourite sites like a social feed? HyperTexting is basically a richer RSS app that tried to present the open web in a way that mirrors the social web. https://techcrunch.com/2026/07/10/a-new-app-hypertexting-turns-the-open-web-into-a-scrollable-social-media-like-feed/
Dave Winer used the Really Simple Syndication (RSS) format to… make a social network? Using nothing "more than RSS 2.0, OPML, Markdown, SQL and WebSocket" he has created rss.chat a simple chat network. http://scripting.com/2026/07/10/161133.html
The story behind Haystack, BBC Eye’s agent system (with human‑in‑the‑loop) used to process 10,000 Russian social media posts. https://reutersinstitute.politics.ox.ac.uk/news/how-bbc-eye-built-multi-agent-ai-system-sift-through-ten-thousand-russian-social-media-posts
Arc XP (The Washington Post’s tech arm) launched Ask The News an embeddable AI bot that keeps queries, data, and monetization on publisher's own sites. The chatbot can trigger subscription meters off AI answers, and integrates contextual ads while preserving control. https://www.arcxp.com/products/ask-the-news/
