Cannes Gets AI Lennon, Congo Gets Ebola, You Get Both

The story that should be leading every newscast

A Bundibugyo strain Ebola outbreak in the Democratic Republic of Congo has produced around eighty suspected deaths, and the world does not have a licensed vaccine for this particular variant. The ring-vaccination playbook that contained recent Zaire-strain outbreaks does not directly apply. Bundibugyo is rare enough that nobody bothered to finish the homework, and now the homework is due.

This is the only story in today's slot that genuinely matters on a human timescale, and it is competing for attention with a film festival and a camera app. That tells you something about the information environment more than it tells you about epidemiology. The rest of the debrief is what the industry wants you looking at instead.

Cannes discovers AI is a competent intern

Steven Soderbergh's John Lennon documentary premiered at Cannes with extensive AI-generated sequences stitched into the archival material. Reviewers describe the synthetic passages as bland, inoffensive, and additive in the way a third helping of mashed potatoes is additive. The interesting finding is not that AI video looks fake. It is that AI video looks fine, and fine is the enemy of documentary filmmaking, which is supposed to feel like something specific happened to a specific person.

This is the actual creative risk that nobody on a panel wants to articulate. Generative tools do not flood the zone with garbage. They flood it with a smooth gray paste of acceptable output, and acceptable output is what kills the form. Soderbergh, of all people, should have seen this coming.

Sony, in a parallel act of caution, spent the day insisting that its new AI Camera Assistant only suggests edits rather than performing them. The suggestion, presumably, is also gray paste. Instagram went the other direction and launched a feature that fires your photos to mutual followers the instant you take them, because the friction of deciding whether to post something was apparently the last bottleneck standing between you and total social transparency. Privacy advocates will object. The feature will ship anyway and most users will leave it on.

Infrastructure meets the people who live near it

The data center fight is becoming a real fight. Communities adjacent to hyperscaler builds are organizing against water draws, grid strain, noise, and the general sense that the local tax base is subsidizing somebody else's chatbot margins. The industry's defensive posture is new. A year ago the standard response was a press release about jobs. Now it is lawyers and community liaisons, which means the opposition is landing punches.

This is the part of the AI buildout that does not show up in benchmark charts. Compute has to sit somewhere, drink something, and exhaust heat into somebody's air. The political economy of where that somewhere is located is going to define the next two years of infrastructure policy more than any model release.

Courtrooms, classrooms, and the labor question

The Musk versus Altman trial went to a jury, which is the legal system's way of admitting that the underlying dispute is unfalsifiable. Did OpenAI betray its charitable mission? Nine people in a San Francisco jury box will now decide, and whatever they decide will be appealed. Treat this as ambient noise until the appellate ruling.

A CEO survey making the rounds suggests AI-driven layoffs may spare older workers, reversing the usual pattern where age correlates with vulnerability. The mechanism, as reported, is that institutional knowledge becomes scarcer and more valuable when entry-level pipelines get automated. Possibly true. Also exactly what executives would say in a survey conducted while they prepare to cut entry-level pipelines.

Finally, a feature on gamified math homework documents children spending more time on the game wrapper than on the math underneath. The product works as designed. The design is the problem. We have built educational tools that optimize for engagement and then act surprised when engagement is what they produce. The Lennon documentary, the Sony camera, and the math game are the same story told three times. Software that aims at adequacy hits adequacy, and adequacy is not what any of these domains needed.