Real breakthroughs don’t always come from following the standard rules, they come from questioning it.While AI giants battle to build bigger models, consume more data, and scale GPU farms, Jesús Manuel Soledad believes the world needs the opposite: smaller, more efficient, more private AI that doesn’t depend on massive infrastructure.

In this episode, Jesús shares how he went from a lawyer in Chihuahua, Mexico to the founder of Marté AI, a company daring to challenge the foundational assumptions of modern AI. By asking the questions no one else is asking—Do we actually need massive datasets? Why must AI depend on GPUs? What if intelligence can run fully offline?—Marté AI is reinventing how AI is built.

We dive into:

• Why Jesús left law to self-learn AI during the pandemic

• How challenging the “more data, bigger models” mantra unlocked new discoveries

• The surprising insights found by training a single neuron from scratch

• Why Marté AI believes the AI industry took a wrong turn years ago

• How offline AI can protect privacy—and why centralizing intelligence is dangerous

• The hybrid symbolic + neural approach enabling models under 5 MB

• Why efficient AI could save massive energy, water, and compute globally

• The philosophical question: Are we discovering intelligence… or reverse-engineering a happy accident?

Follow ExitFund for more founder stories from the frontier of technology. Share this episode with someone who believes true innovation starts by asking, “Says who?”