AI News Feed
OpenAI Roadmap: Autonomous Research Agents by 2028
01 Nov 2025- OpenAI unveiled a multi-year plan to build autonomous research agents—intern-level by 2026, fully automated by 2028—backed by massive compute, funding, and organizational restructuring.
01 Nov 2025
OpenAI announced a multi-year roadmap to build autonomous research agents: an “intern‑level” research assistant by September 2026 and a “fully automated, legitimate AI researcher” by 2028, Sam Altman said on a livestream reported by The Neuron. Chief Scientist Jakub Pachocki described the target as systems that can run larger research projects end‑to‑end — form hypotheses, run experiments, analyze results, and iterate without human hand‑holding.
The plan leans on three pillars: continued algorithmic advances, dramatically more “test‑time compute” (potentially dedicating substantial datacenter resources to single problems), and much longer model time horizons (current models handle roughly five‑hour windows today). The newsletter also notes organizational and financial moves: OpenAI completed a shift to a public‑benefit structure where a non‑profit foundation will own 26% of the for‑profit arm and commit $25B toward disease research, while the for‑profit side can raise capital for large infrastructure builds. Altman referenced a 30‑gigawatt buildout and characterized the effort as a multiyear, multitrillion‑dollar scale commitment.
If OpenAI hits these milestones, the newsletter argues, AI could materially accelerate scientific discovery — and Pachocki warned deep‑learning systems might be “less than a decade away from superintelligence.”
Source
OpenAI announced a multi-year roadmap to build autonomous research agents: an “intern‑level” research assistant by September 2026 and a “fully automated, legitimate AI researcher” by 2028, Sam Altman said on a livestream reported by The Neuron. Chief Scientist Jakub Pachocki described the target as systems that can run larger research projects end‑to‑end — form hypotheses, run experiments, analyze results, and iterate without human hand‑holding.
The plan leans on three pillars: continued algorithmic advances, dramatically more “test‑time compute” (potentially dedicating substantial datacenter resources to single problems), and much longer model time horizons (current models handle roughly five‑hour windows today). The newsletter also notes organizational and financial moves: OpenAI completed a shift to a public‑benefit structure where a non‑profit foundation will own 26% of the for‑profit arm and commit $25B toward disease research, while the for‑profit side can raise capital for large infrastructure builds. Altman referenced a 30‑gigawatt buildout and characterized the effort as a multiyear, multitrillion‑dollar scale commitment.
If OpenAI hits these milestones, the newsletter argues, AI could materially accelerate scientific discovery — and Pachocki warned deep‑learning systems might be “less than a decade away from superintelligence.”
Source
The method
The prompts
Helping entrepreneurs and business owners use AI tools for growth and productivity.
Wawa Llama Ltd GB447340593