🤖AI Brain

Your panel broadly agrees that frontier capability keeps improving and that task-level automation is already real. They split hard on how quickly that converts into full role replacement and durable enterprise revenue.

The themes your panel returned to most across 148 supported sources.

• How much of today's AI capex converts into durable earnings
• Task-level automation is already measurable in support, translation and routine analysis
• Whether current architectures reach reliable long-horizon reasoning
• Model capability is still improving faster than organisations can absorb it

• Task-level automation is already measurable in support, translation and routine analysis
• Compute and energy availability are the binding constraints through 2027
• Model capability is still improving faster than organisations can absorb it

Disagreement is preserved, never averaged away.

• Whether current architectures reach reliable long-horizon reasoning
• How much of today's AI capex converts into durable earnings
• Speed of full role replacement outside routine work

• 2 hours ago — Two new lab statements ingested; agreement +3 pts
• This week — Skeptic camp softened on coding agents
• Last month — Capex-return debate re-opened after earnings

• 8 new sources entered the record this period.
• Two new lab statements ingested; agreement +3 pts
• Agreement is at 72% across 15 tracked experts.

• Jensen Huang

  Reiterated their position on task-level automation is already measurable in support, translation and routine analysis.

• Sam Altman

  Pushed back on how much of today's ai capex converts into durable earnings.

• Dario Amodei

  Added new material this period; position unchanged.

• Demis Hassabis

  Reiterated their position on task-level automation is already measurable in support, translation and routine analysis.

• Mark Zuckerberg

  Pushed back on how much of today's ai capex converts into durable earnings.

• Satya Nadella

  Added new material this period; position unchanged.

• Andrej Karpathy

  Reiterated their position on task-level automation is already measurable in support, translation and routine analysis.

• Fei-Fei Li

  Pushed back on how much of today's ai capex converts into durable earnings.

• How much of today's AI capex converts into durable earnings39
• Task-level automation is already measurable in support, translation and routine analysis38
• Whether current architectures reach reliable long-horizon reasoning38
• Model capability is still improving faster than organisations can absorb it25
• Speed of full role replacement outside routine work14
• Compute and energy availability are the binding constraints through 20279

The minority position worth taking seriously: Speed of full role replacement outside routine work. It is held by a small part of your panel but is supported by evidence in the record.

• Is NVIDIA still the strongest AI stock?
• Which AI companies have the highest conviction?
• What changed after earnings?
• Which experts became more bullish?

Earnings calls & filings · Frontier lab technical reports · Long-form podcast interviews · Peer-reviewed benchmarks