Teahose.
SIGN IN
NEW HERE — WHAT TEAHOSE DOES
We read the entire AI & tech firehose — so you don't have to.
PODPodcastsAll-In, No Priors, Acquired…
NEWNewslettersStratechery, Newcomer…
PAPPapersPhysical AI research
PHProduct Huntdaily launches
VCInvestor ScoutSequoia, a16z, Benchmark…
CLAUDE DISTILLS →
7 reads, 30 sec each — free, 6 AM ET.
+ a live graph of the companies, people & themes underneath.
HOME/OUR WORLD IN DATA/The OWID Brief: Child deaths, LG…
NEWS
// NEWSLETTER ISSUE
OUR WORLD IN DATA

The OWID Brief: Child deaths, LGBT+ rights, electric vehicle sales, and more

DATE June 5, 2026SOURCE OUR WORLD IN DATAPARTICIPANTS OUR WORLD IN DATA
// SUMMARY

1. Key Themes


Electric Vehicles Are Crossing a Mass-Market Inflection Point

Global combustion engine car sales peaked in 2017, and by 2025, one in four cars sold was electric. These two data points together signal that the EV transition is no longer a future event — it is structurally underway.

"Global sales of combustion engine cars peaked in 2017" and "One in four cars sold in 2025 was electric."


AI Is Transitioning From Tool to Autonomous Research Agent

OpenAI's general reasoning model independently disproved a conjecture that stumped mathematicians for 80 years — without being specially trained on the problem. This signals a step-change in AI's role across science and industry.

"This marks the first time an AI has autonomously resolved a prominent open question central to a mathematical field… The AI model was not specially trained on math or this problem in particular, but is a general reasoning model. Those general capabilities could make it a useful research partner beyond math, such as in physics, engineering, biology, and medicine."


Global Health Inequality Is an Addressable Market Gap, Not an Intractable Problem

Five million children die annually, but the wide variation in mortality rates across countries implies the solutions already exist — the gap is in deployment, not discovery.

"There are huge gaps in child mortality rates between countries. This is good news: it means we already know how to prevent most of these five million children dying."


Democratic Institutions Are More Resilient Than the Prevailing Narrative Suggests

Researchers studied 27 cases of democratic erosion since 1999 and found that in three-quarters of cases, would-be autocrats were stopped — most often by elections and term limits themselves.

"In three-quarters of cases, would-be autocrats were stopped short of fully eroding democracy in their country. Most often, they were succeeded by new leaders who halted or even reversed the erosion."



2. Contrarian Perspectives


Voting Is Actually One of the Most Effective Tools Against Autocracy The conventional wisdom in activist circles is that electoral politics is insufficient to defend democracy. The research evidence pushes back sharply on this.

"The common advice to 'go vote' to defend democracy can sound trite… But recent history shows that even increasingly unfair elections can be won if enough people turn out." The study of 27 real-world cases found elections and term limits were the most common mechanism for stopping democratic erosion — not protests, courts, or international pressure.


The Child Mortality Crisis Is Not Primarily a Knowledge Problem It might seem intuitive that deaths persist because we lack medical solutions. The data inverts this: the bottleneck is application of known interventions, not R&D.

"Most of these deaths are preventable… it means we already know how to prevent most of these five million children dying. The first step is knowing what they're dying from." The geographic disparity across income levels — where infectious diseases dominate in poor countries but are "far rarer" in rich ones — confirms the gap is systemic, not scientific.


General-Purpose AI May Outperform Domain-Specific AI on Hard Research Problems The expectation might be that breakthroughs in math or science require narrowly trained specialist models. The OpenAI result challenges this assumption.

"The AI model was not specially trained on math or this problem in particular, but is a general reasoning model." If general reasoning capability is sufficient to crack 80-year-old open problems, the investment thesis favoring narrow vertical AI models in research applications deserves scrutiny.



3. Companies Identified

CompanyDescriptionWhy MentionedQuote
OpenAIAI research companyIts internal general reasoning model autonomously disproved the Erdős planar unit distance conjecture — a landmark first in AI-driven mathematical research"OpenAI recently announced that an internal model, working entirely on its own, had made a breakthrough on the problem by disproving Erdős's conjecture and offering a better construction."


4. People Identified

PersonDescriptionWhy MentionedQuote
Hannah RitchieResearcher/writer at Our World in DataAuthored the analysis on child mortality causes by country and income level"Hannah Ritchie wrote an article that uses the visualization to walk you through the data."
Sophia MersmannResearcher at Our World in DataCo-built the interactive child mortality visualization"Our colleagues Sophia Mersmann and Fiona Spooner built a new interactive visualization."
Fiona SpoonerResearcher at Our World in DataCo-built the interactive child mortality visualization"Our colleagues Sophia Mersmann and Fiona Spooner built a new interactive visualization."
Paul ErdősLegendary mathematician (1913–1996)Posed the planar unit distance problem in 1946; his conjecture was disproved 80 years later by an AI"In 1946, mathematician Paul Erdős posed a now-famous problem in discrete geometry… He put forward his own conjecture about the solution, and offered a monetary prize to anyone who could resolve it. For 80 years, no one could."
Tim GowersFields Medal-winning mathematicianValidated the AI's mathematical breakthrough, lending it the highest credibility in the field"Tim Gowers, a Fields medalist (math's highest honor), called it 'a milestone in AI mathematics'."
Kristopher VelascoResearcher at Princeton UniversityCompiled the LGBTI National Policy Dataset tracking 27 policy dimensions across 197 countries since 1991"This data comes from the LGBTI National Policy Dataset, compiled by researcher Kristopher Velasco at Princeton University."
Javier CorralesDemocracy researcherCo-authored study of 27 democratic erosion cases since 1999; findings showed autocrats were stopped in 75% of instances"Democracy researchers Javier Corrales and Susan Stokes identified and studied 27 cases of democratic erosion since 1999."
Susan StokesDemocracy researcherCo-authored the Journal of Democracy study on resisting autocratic erosion"Their findings, shared in a recent article in the Journal of Democracy, are encouraging."


5. Operating Insights


Disaggregate your data before designing interventions. The child mortality analysis shows that a single global statistic obscures vastly different root causes across income groups. In rich countries, non-communicable diseases and injuries dominate; in poor countries, infectious diseases dominate. Applying a one-size-fits-all solution would misallocate resources.

"This picture looks very different across income levels. In poorer countries, infectious diseases are even more prominent. In rich countries, they are far rarer, and non-communicable diseases and injuries account for a larger share." Operator takeaway: Before deploying capital or programs at scale, segment your problem by context — the leading cause of failure in your market may not match the global average.


Treat general-purpose AI as a research partner now, not a future consideration. The OpenAI result demonstrates that frontier general reasoning models can generate novel, expert-validated breakthroughs autonomously, without domain-specific fine-tuning.

"Those general capabilities could make it a useful research partner beyond math, such as in physics, engineering, biology, and medicine, potentially accelerating progress in those fields." Operator takeaway: Entrepreneurs in research-intensive sectors (biotech, materials science, climate tech) should be actively testing whether general reasoning models can compress their discovery timelines.



6. Overlooked Insights


Coffee production geography has quietly shifted over six decades. The newsletter flags that "coffee production has shifted toward Asia over the last six decades" — a single line with significant implications for commodity traders, agricultural investors, and consumer brands dependent on traditional Latin American or African supply chains. A decades-long geographic realignment in a globally consumed commodity is the kind of slow-moving structural shift that tends to be underpriced until it isn't.


Waste infrastructure in low- and middle-income countries is a large, underserved market. The newsletter notes that "most collected waste in many low- and middle-income countries is stored in open dumps or is burned." This is both an environmental crisis and a significant infrastructure gap — representing a potential investment opportunity in waste management, recycling technology, and circular economy businesses targeting emerging markets, where the baseline is extremely low and regulatory pressure to improve is likely to grow.