The structural paradox of Jekyll and Hyde

My Nieman Labs piece (“Google Will Look Beyond Volume Journalism”) sparked a vital debate. Much of the feedback focused on the perceived tension between editorial “soul” and engineering “efficiency.”

In this post, I want to argue that the engineering vs. editorial rhetorical horse has been beaten to death; it is time we look at the actual physics of the problem.

The systemic trap

This “split personality” is not a choice; it is an industry-wide survival mechanism.

By day (Mr. Hyde): The industry bows to the Algorithm God. This is the side of the business driven by the cold reality of the “volume trap”—the systemic need to commission the 500th definitive guide to “Chia Seeds” or chase trending keywords to sustain digital reach. It is a high-volume, high-velocity mandate that treats information as a commodity.

By night (Dr. Jekyll): This is the editorial soul emerging to speak of Pulitzer-worthy exposés and to criticize technologists for taking complex, high-friction realities—like philosophy, human behavior, and the soul of journalism—and strip-mining them into simplified, data-friendly units.

By day (Mr. Hyde): The industry knows that most digital revenue comes from the Algorithm God—a deity that weighs investigative journalism and funny cat videos on the exact same scale (eCPM & impressions).

By night (Dr. Jekyll): They long to write in a deeply lyrical voice—a stylistic luxury that the economics of volume would never allow the volume newsroom to indulge in.

But here is the reality: Users do not change their species when they jump platforms. We do. The tragedy of the last decade is that these two personas have been forced to live separate lives. The “Hyde” of volume treats depth and single-peg stories as interchangeable because the Algorithm God weighed them on the exact same scale (eCPM & impressions). This has eroded the very trust we seek to protect.

I am a nine-year newsroom native and two-decade technologist. In transitioning from leading Editorial Product to Machine Learning, I didn’t abandon “what got us here”; I am building “what will take us there.” To ignore the shifting physics of the internet isn’t native wisdom—it is a failure to see the horizon.

Tangentially: 

Also Read: Apart from my work with Machine Learning and Generative AI, I am looking to discover what ‘AI Irreducible Journalism’ would look like. Here’s my latest attempt: Inside Look: How Jewelers Recycle India’s Family Heirlooms.

The shared reality

In true algorithmic style, most criticism of algorithms can be reduced down to one word: “Reductive.”

Yes, it is true. Algorithms are reductive. In ML, we call this “Dimensionality Reduction.” This industry prides itself on reducing nuance down to brass tacks. Dimensionality reduction isn’t about losing nuance—it’s about building algorithms that can run the internet within the given compute and latency (GPU and RAM) constraints.

But let’s be clear: Editorial Judgment and journalism itself are “Dimensionality Reduction”:

• Editorial judgment is choosing not to cover certain aspects of reality so the article fits in neat narrative arcs.
• A headline is a 1D reduction of a 3D event—an essential loss of nuance to gain the reader’s attention.
• A 300-word summary is a dimensionality reduction of a 10-page report.

The problem isn’t reducing… it is reducing the right thing! Until now, the algorithm reduced articles down to keywords. Now, it can reduce them down to new measures, and with them, user behavior on the internet will change.

It is what it is

What is also true is that most of the internet and digital economy is governed by algorithms and algorithm people. Therefore, we must observe and forecast how the algorithmic internet will pivot.

In this transition, the risk is not in being “too technical” or “too reductive.” The risk is in being too slow. As Sundar Pichai noted, “It’s better to overinvest than to underinvest in AI.”

A work-in-progress thesis for the coming decade

The observations shared on MediaFlywheels are an honest attempt to put theoretical frameworks out in the open and invite constructive debate to meaningfully compete with algorithms in the coming decade. The hope is that the push and pull of open debate will move the industry needle forward.

This is what we technologists call #BuildInPublic. This is what Taleb calls Skin In The Game. And sharing these frameworks on LinkedIn allows for a peer-review process among the global news intelligentsia.

What’s algorithmic oxygen

Much like cash flow is the oxygen of a business, increasingly information (tokens)—and not articles (web links)—is likely to become the commodity that will be traded with the internet’s Algorithmic Gods.

In the coming years, the industry objective will shift: Maximize OpenAI Licensing Fees While Growing O&O Audience.

And with it, much like editors drive “Content Strategy,” newsrooms globally are likely to require a “Token Strategy” in coming years. This is where new industry metrics come in:

• AI Irreducibility Score
• Net New Information Rate
• Information Latency Rate 

The hope

The tragedy of the last decade or two was that the Algorithm God demanded Mr. Hyde; the opportunity of the next decade is that the algorithmic pivot toward LLMs finally allows us to let Dr. Jekyll lead again.

In fact, the more I think about it… “Old school journalism is AI irreducible.” Hence, I am advocating for a sector-wide doubling down on news gathering (the input side) and moving away from the commoditized desk.

The measured test 

Since 2023, I’ve advocated for separating content and product strategies for owned (logged-in web and app) and rented (logged-out web & YouTube) audiences, though I didn’t begin writing about the topic formally until 2024.

This is a logical pivot:

• On all commodity information, follow the trend fast and let product short-circuit Google’s algorithmic advantage by adopting personalization, topic modeling, AI summarization, and AI writing of articles.
• With that out of the way, let “old school journalists” compete on ‘hand-crafted’ journalism that BigTech models cannot compete on.

What next

I will address the remaining 60% of the structural criticism in the second version of my Nieman Labs analysis. My focus will be on bridging the gap between editorial intuition and algorithmic reality for a new global economic reality.

Happy Gregorian New Year.

END OF ARTICLE