The Download: NASA’s new space telescope and OpenAI’s autonomous hacker

Drafting five analytical paragraphs from only the facts you provided, then logging the task.MIT Technology Review’s weekday newsletter The Download led with two stories: NASA’s Nancy Grace Roman Space Telescope, which may launch as early as the end of next month, and OpenAI’s autonomous hacker. On the space side, the piece highlights shape-shifting mirrors on Roman and the claim that those optics could help unveil Jupiter-like planets around other stars—worlds more like our own system’s gas giant than the hot, close-in giants that dominate many current detections.

The technical hook in the briefing is the mirrors themselves. Rather than fixed optics, Roman is framed around shape-shifting mirrors that can adjust in flight—an active-control approach meant to keep the optical path stable enough for the faint signals of distant planets. The Download does not publish bench numbers here; the engineering point is adaptive figure control as a flight system, not a ground lab demo, timed to a launch window that could open by the end of next month.

For builders, that is a concrete systems problem: closed-loop sensing, actuators, and thermal stability under launch and on-orbit conditions. The same control-and-metrology mindset shows up in high-precision instruments, deformable optics, and any product where mechanical or thermal drift would kill the signal. The OpenAI autonomous-hacker item sits in the same issue as a parallel systems story—software that acts with less human step-through—so the shared theme is autonomy under real constraints, not a single product pitch.

In market and program terms, Roman is a flagship NASA astronomy mission competing for attention with other large space telescopes and ground surveys that already hunt exoplanets by transit, radial velocity, or direct imaging. Shape-shifting mirrors are the differentiator The Download foregrounds for Roman’s path to Jupiter analogs. On the AI side, an autonomous hacker from OpenAI lands in a crowded field of offensive-security tools, red-team automation, and agentic systems, where capability claims are judged against reliability, scope, and who is allowed to run them.

What to watch next is binary and near-term: whether Roman holds a launch as early as the end of next month, and whether post-launch reporting confirms that the shape-shifting mirrors perform as described for Jupiter-like planet work. On the OpenAI track, watch how The Download and follow-on coverage define what “autonomous” means in practice—task boundaries, human oversight, and where the system stops—rather than the label alone.MIT Technology Review’s weekday newsletter The Download led with two stories: NASA’s Nancy Grace Roman Space Telescope, which may launch as early as the end of next month, and OpenAI’s autonomous hacker. On the space side, the piece highlights shape-shifting mirrors on Roman and the claim that those optics could help unveil Jupiter-like planets around other stars—worlds more like our own system’s gas giant than the hot, close-in giants that dominate many current detections.

The technical hook in the briefing is the mirrors themselves. Rather than fixed optics, Roman is framed around shape-shifting mirrors that can adjust in flight—an active-control approach meant to keep the optical path stable enough for the faint signals of distant planets. The Download does not publish bench numbers here; the engineering point is adaptive figure control as a flight system, not a ground lab demo, timed to a launch window that could open by the end of next month.

For builders, that is a concrete systems problem: closed-loop sensing, actuators, and thermal stability under launch and on-orbit conditions. The same control-and-metrology mindset shows up in high-precision instruments, deformable optics, and any product where mechanical or thermal drift would kill the signal. The OpenAI autonomous-hacker item sits in the same issue as a parallel systems story—software that acts with less human step-through—so the shared theme is autonomy under real constraints, not a single product pitch.

In market and program terms, Roman is a flagship NASA astronomy mission competing for attention with other large space telescopes and ground surveys that already hunt exoplanets by transit, radial velocity, or direct imaging. Shape-shifting mirrors are the differentiator The Download foregrounds for Roman’s path to Jupiter analogs. On the AI side, an autonomous hacker from OpenAI lands in a crowded field of offensive-security tools, red-team automation, and agentic systems, where capability claims are judged against reliability, scope, and who is allowed to run them.

What to watch next is binary and near-term: whether Roman holds a launch as early as the end of next month, and whether post-launch reporting confirms that the shape-shifting mirrors perform as described for Jupiter-like planet work. On the OpenAI track, watch how The Download and follow-on coverage define what “autonomous” means in practice—task boundaries, human oversight, and where the system stops—rather than the label alone.