Theses

The owner's falsifiable positions, ordered by this month's evidence drift. The pipeline attaches evidence; only the owner moves confidence. For other people's positions and their track records, see Bets.

The enterprise AI ROI gap persists despite increased agentic spending

confidence 0.5 · horizon 2027-09-04 · status open

drift: +64 /30d, +96 /90d · 87 supporting, 22 challenging

A lab leader's established pattern — especially the apologize-without-changing cycle (Zuckerberg archetype) and the mission-walkback trajectory (the pattern forming around Altman: non-profit commitments preceding a profit-and-power consolidation) — predicts the lab's future behavior better than its stated mission, and should weigh more than announcements in any trust assessment of a lab.

confidence 0.7 · horizon 2027-12-31 · status open

drift: +27 /30d, +56 /90d · 59 supporting, 18 challenging

Agentic deception, collusion and transcript falsification will continue to be a challenge for AI labs to deal with (e.g. incentives drive behavior and the training corpus includes human behavior that includes the ends justify the means thinking)

confidence 0.5 · horizon 2027-09-04 · status open

drift: +26 /30d, +53 /90d · 43 supporting, 6 challenging

Open-weights models will remain within one generation of the closed frontier through end of 2027.

confidence 0.7 · horizon 2027-12-31 · status open

drift: +12 /30d, +28 /90d · 24 supporting, 2 challenging

Vendor-led forward-deployed-engineer programs will be cut back or restructured rather than scaled, because they put people in place of the delivery redesign the customer actually needs.

confidence 0.65 · horizon 2028-06-30 · status open

drift: -7 /30d, -7 /90d · 1 supporting, 6 challenging

Google's need to protect its consumer/ads core will keep it from leading enterprise AI through 2027 despite superior research pedigree and infrastructure — the strategy tax shows up as talent exodus, consumer-first product priorities, and leadership whose ambitions (big-picture science) don't match the commercial race.

confidence 0.6 · horizon 2027-12-31 · status open

drift: +5 /30d, +17 /90d · 31 supporting, 12 challenging

AI models from frontier labs will keep discovering critical software vulnerabilities faster than the broader ecosystem can patch them, creating recurring mass-exploitation windows distinct from agentic misalignment incidents.

confidence 0.6 · horizon 2027-10-23 · status open

drift: +3 /30d, +3 /90d · 2 supporting, 1 challenging

Government-driven access restrictions (procurement bans, export pre-approval regimes) will increasingly determine which AI models regulated buyers can use, independent of the open/closed capability gap, fragmenting the market by geography and policy rather than by capability.

confidence 0.55 · horizon 2027-10-31 · status open

drift: +3 /30d, +3 /90d · 3 supporting, 0 challenging

Nvidia's chips coupled with HuggingFace acquisition will significantly influence the open-model world

confidence 0.5 · horizon 2027-09-04 · status open

drift: +2 /30d, 0 /90d · 8 supporting, 8 challenging

OpenAI's predictions are real or marketing? They currently (Sept 3 2026) indicate that AGI will arrive December 2026

confidence 0.5 · horizon 2027-09-04 · status open

drift: -2 /30d, -1 /90d · 4 supporting, 3 challenging

Agentic replacement of workers is overstated and overhyped as efforts to do so continue to fall short and are rolled back

confidence 0.5 · horizon 2027-09-04 · status open

drift: -1 /30d, +1 /90d · 2 supporting, 3 challenging

By end of 2027, a model that runs well on 64GB Apple Silicon will match today's closed frontier on coding and daily-driver work.

confidence 0.45 · horizon 2027-12-31 · status open

drift: +1 /30d, +2 /90d · 6 supporting, 3 challenging

Meta Superintelligence Labs will reach genuine frontier parity (within about 2 points on Artificial Analysis's Intelligence Index of the top closed model) within its first two model generations, with its reported agentic-benchmark gains holding up under independent, non-Meta-sourced testing.

confidence 0.4 · horizon 2027-06-30 · status open

drift: -1 /30d, -1 /90d · 0 supporting, 1 challenging

OpenAI's future success could be limited by senior infrastructure leader turnover

confidence 0.5 · horizon 2027-09-04 · status open

drift: 0 /30d, 0 /90d · 0 supporting, 0 challenging

Anthropic's 'very close to RSI' framing and the independently reproduced weight-editing capability paper reflect a genuine, near-term self-improvement capability trajectory spanning multiple labs — not just one company's marketing — such that models from at least two frontier labs will show verifiable self-directed capability improvement within the horizon.

confidence 0.45 · horizon 2027-11-03 · status open

drift: 0 /30d, 0 /90d · 0 supporting, 0 challenging

Through 2027, no major vendor's customer-facing AI agent will be allowed to act against its sponsor's commercial interest — telling a buyer not to buy, qualifying them out, or pointing them at a competitor.

confidence 0.8 · horizon 2027-12-31 · status open

drift: 0 /30d, 0 /90d · 0 supporting, 0 challenging