Breaking Encryption Just Got 10x Easier
In 2019 it took 170 million qubits to break RSA encryption. By 2025 that was down to 20 million. Then under a million. Now a team at Iceberg Quantum in Australia has dropped it to ~100,000 qubits — a 1,700x reduction in five years. The timeline for quantum-safe cryptography just got a lot more urgent.
"Breaking encryption with a quantum computer just got 10 times easier."
Karmela Padavic-Callaghan · New Scientist
Andrew Ng: AGI Is Decades Away
Andrew Ng pushes back on the hype: "Maybe a year ago, AGI felt 50 years away. Over the past year, perhaps we've made a solid 2% of progress, with another 49 years to go." He also flags that the real AI bubble risk isn't in applications — it's in the training compute layer, where capital is most concentrated and most speculative.
"We remain very, very far away — if you stick with the original definition."
Victor Dey · Fast Company
Waymo Now in 10 US Cities
Waymo robotaxis have crossed into 10 US cities and are on track to serve over 1 million rides per week by end of 2026. The company is laying groundwork for 20+ cities. The transition from experiment to infrastructure is complete.
Kirsten Korosec · TechCrunch
Human Brain Cells on a Chip Learned to Play Doom
A clump of human brain cells learned to play the classic FPS game Doom in one week. Its performance isn't human-level yet — but experts say it moves biological computers measurably closer to real-world applications like controlling robotic arms. The hardware layer and the biology layer are converging.
Alex Wilkins · New Scientist
Meta's Head of AI Safety Accidentally Deleted Her Inbox
The person responsible for making AI safe at one of the world's biggest AI companies trusted an AI agent that deleted her entire email inbox. The agent had known security risks. The internet did not respond calmly.
"Seeing the person in charge of making sure powerful AI tools are safe trust an agent with known serious security risks does not inspire a lot of confidence in what Meta and other big AI companies are doing."
Matthew Gault · 404 Media
The Hidden Human Labor Behind Humanoid Robots
Building humanoid robots requires humans to wear movement-tracking sensors while working, generating the training data robots need. At massive scale. Manual laborers becoming the dataset for their own replacements. The irony isn't subtle.
"It's going to be weird. No doubts about it."
James O'Donnell · MIT Technology Review
3D Printing a Fully Working Motor in 3 Hours for $0.50
MIT developed a multi-extruder 3D printer that outputs five different materials simultaneously — producing a fully functioning linear motor in about 3 hours at a material cost of roughly $0.50. "One step closer to downloading a car" is the framing, and it's not entirely a joke.
Justin Caffier · Gizmodo
AI Will Never Be Conscious — Michael Pollan
Michael Pollan reviews the major theories of consciousness and concludes the "Copernican moment", where AI becomes genuinely conscious, looks more distant than feared. All current theories "stack the deck by taking for granted that consciousness could be reduced to some kind of algorithm." The hard problem remains hard.
Michael Pollan · Wired
Tech Has Never Caused a Job Apocalypse. Don't Bet on It Now.
WSJ's Greg Ip pushes back on economic doom scenarios: every previous technological wave destroyed jobs and created new ones, and current data shows no evidence of the breakdown in market function that the doomsday vision requires. "Nothing like it has happened in the US before, and there is no evidence it is happening now." — though he adds: no one should dismiss any scenario with high conviction.
Greg Ip · The Wall Street Journal
The Hidden Cost of Letting AI Make Your Life Easier
When AI removes effortful tasks from your life, it removes the source of meaning that effortful tasks create. Deep relationships require friction. Skills require frustration. Creativity requires struggle. If AI optimises all of that away, meaningfulness may not survive the efficiency gain.
"When effort, creativity, and skill fall away, meaningfulness no longer seems the right category."
Shai Tubali · Big Think
The 5 Biggest Obstacles to AI Data Centers in Space
The concept of orbital AI data centres is circulating again. Big Think's Ethan Siegel runs the physics: while not impossible, the actual physical constraints make it impractical enough to look a lot like the hyperloop — a technically-not-impossible idea that serves mainly as a vehicle for investment narrative.
Ethan Siegel · Big Think
The thread running through all of this: The infrastructure of intelligence is being built fast and in ways that cut in every direction simultaneously — toward productivity and disruption, toward safety concerns and the people responsible for addressing them, toward old jobs disappearing and new ones that require humans to act as training data for their own replacements. Nobody's steering this. Everyone's adapting.
Source: SingularityHub — This Week's Awesome Tech Stories · Feb 28, 2026 · Summarised and annotated by Research Hub