Everyone’s talking about the math problem an AI supposedly solved this week, but almost no one’s looking at the real money that was poured into it. Meanwhile, the money that’s actually verifiable is already moving — and it’s in Chocó, Colombia. Let’s start there.
Intelligence Pulse
The “Plan Marshall” President Abelardo de la Espriella presented on September 7 in Quibdó isn’t just another post-earthquake reconstruction announcement: it’s the fifth time in three decades a Colombian government has put an interoceanic corridor for Chocó on the table — and this time it comes with real capital already committed. Colombia’s private sector pledged 2 trillion pesos (roughly US$500 million) for reconstruction, the state announced an additional 1 trillion pesos for roads, an airport and connectivity (split evenly between tax-based public works and private funding), and 5 billion pesos in seed capital via iNNpulsa is already reaching close to 1,000 local entrepreneurs, backed by Colombina, Bancolombia and Tecnoglass. The megaproject itself — rail or canal — remains, by track record, a high execution-risk bet: no previous government has ever completed it. But the capital already flowing into logistics, construction materials, tourism and seed funding is a concrete, near-term opportunity — regardless of whether the corridor itself gets built this decade or the next.
AI Frontier
OpenAI announced that an internal model — more capable than its recently launched GPT-6 Astra — solved the Navier-Stokes existence and smoothness problem, one of the six Clay Mathematics Institute Millennium Problems. The number that actually matters isn’t the equation: it’s the 10,000 coordinated AI agents that ran for 88 hours, exchanging 2.7 million messages, at a cost its own researcher, Sébastien Bubeck, estimated at “several million dollars”. Here’s what to take from that: solving complex problems no longer depends solely on talent — it depends on how much compute you can buy. For any Colombian company using Codex, Claude or Copilot in daily operations, there’s a more urgent second takeaway: OpenAI admitted that “de-identified” data from its products may have helped improve its models. Check your AI vendor’s terms of use before feeding it sensitive business information.
Signal vs. Noise
The noise: headlines said OpenAI “solved millennium math,” as if it were a closed, unanimous fact. The signal: the result hasn’t cleared peer review yet, OpenAI itself chose not to claim the Clay Institute’s US$1 million Millennium Prize, and two researchers from Harvard and NYU — one of them an Anthropic employee — are accusing the company of possibly benefiting from their private Codex sessions while they worked on the same problem. The market isn’t as convinced as the headlines either: per mentions tracked on Reddit and FinTwit today, neither GeoPark nor any AI ticker beyond the usual names (NVDA, AMD, META) showed a real spike in chatter tied to this news, and overall market sentiment remains neutral. When the price doesn’t move but the headline does, the headline is usually ahead of itself.
Quote of the Day
“Someone’s sitting in the shade today because someone planted a tree a long time ago.” — Warren Buffett
Chocó’s Plan Marshall is a thirty-year bet on shade. This week’s question isn’t whether the corridor will get built — it’s whether it’s worth helping plant the tree, or simply capturing the capital already moving around it.
Want this read every morning before markets open? Subscribe to the ViSFI Daily Briefing.