From the archive
Page 14 of 36 · 842 stories
From the archive
313–336 of 842 articles
SpaceX Launched 170 Rockets in 2025 and Charged for 43. The IPO Filing Shows Where the Other $2.4 Billion Went.
SpaceX's S-1 reveals three numbers that challenge the bull case: ARPU falling to $66/month, xAI burning $1.44 for every dollar Starlink earns, and a valuation of $218,000 per subscriber.
California Let One Professor Cite Herself 76 Times in a 1,000-Page Math Framework. Then the State Board Passed It Anyway.
A Stanford mathematician read every citation in California's math framework. The papers said the opposite of what the framework claimed. We're open-sourcing the tools to audit yours.
A $36 Million Fusion Startup Just Demonstrated What a $15.5 Billion One Hasn't
Realta Fusion generated electricity from fusion plasma using direct energy conversion. The Q-factor math shows why this matters more than the lightbulb it powered.
Two Generic Drugs That Cost $1 a Day Just Showed the Biggest Survival Benefit in ALS History
PrimeC combines celecoxib and ciprofloxacin into an ALS therapy with a 15-month median survival benefit. An original cost-per-survival-month analysis across every ALS drug reveals a potential 160-fold advantage.
A Single Injection Reversed Muscle Loss in a Disease With Zero Treatments. It Didn't Cut Any DNA.
Epicrispr's EPI-321 is the first epigenetic editing therapy to show muscle mass gain in FSHD patients. An original analysis compares epigenetic editing vs DNA-cutting approaches across cost, safety, and scalability.
A 4-Week Drug Restored Aged Stem Cells Without Fixing What Made Them Old. Their Telomeres Were Still Short.
Telomeric antisense oligonucleotides silenced the damage alarm at short telomeres in old mice without lengthening them. Stem cells competed with young ones. Vaccine responses improved. An original cost analysis reveals the economics.
Britain Cancelled Its Next Destroyer to Build Robot Warships Instead. For the Price of One Type 83, the Royal Navy Could Field 20 Uncrewed Missile Barges.
The UK scrapped the Type 83 destroyer and announced four new classes of uncrewed warships within a £5 billion drone investment. An original cost-per-sensor-day analysis shows the hybrid fleet could deliver 7 to 13 times more patrol coverage per pound.
IBM Packs 100 Billion Transistors on a Fingernail. Its Last Breakthrough, From 2021, Still Isn't in Mass Production.
IBM has demonstrated every major chip node first since 2015 and manufactured zero. A systematic analysis of the widening lab-to-factory gap shows why the 0.7nm nanostack may not reach production until 2031.
Tesla Promised 100 GWh of Batteries by 2023. It Has 18 GWh in 2026. Now It Wants to Build a $119 Billion Chip Fab.
Applying Tesla's Battery Day execution trajectory to Terafab economics shows break-even requires 70%+ fab utilization on Intel's unproven 14A node.
Scientists Found 20 Superconductors by Prediction in 60 Years. Machine Learning Could Find Thousands in a Decade.
Two superconductor breakthroughs in 48 hours. ML-guided kagome lattice discovery at Aalto. Magnetic-field-boosted graphene states at MIT. An original energy analysis shows what's at stake.
China Has 15,000 Humanoid Robots Generating Free Training Data. America Built a 90,000-Sq-Ft Facility to Generate Better Data.
AGIBOT shipped 15,000 robots. Apptronik built Robot Park with Google DeepMind. An original cost analysis shows the real bottleneck in humanoid robotics isn't hardware.
Your AI Agent Remembers a Lie It Was Told 3 Months Ago. Researchers Found a 50% Attack Success Rate.
Prompt injection was a session-scoped problem. Sleeper memory poisoning makes it permanent. Three independent research teams found that AI agents with long-term memory can be corrupted once and stay corrupted forever.
CRISPR Just Cured the Youngest Children Ever Treated — Then the Chemo Killed One
Exa-cel achieves 100% efficacy in children ages 5–11 with SCD and beta thalassemia. One child died from busulfan conditioning. The cost math still favors earlier treatment.
The $30 Billion Fuel Cell Boom Has a 60-Tonne Bottleneck
Fuel cell revenue is projected to hit $30B by 2030, up from $2.8B. One company holds virtually every contract. One mineral limits its supply. One country controls that mineral.
A Protein Shell With Zero Enzymatic Activity Became a Working Catalyst. It Cost $1,350 in Lab Work.
Bayreuth's CANVAS workflow transformed inert de novo TIM barrel scaffolds into functional enzymes at 1,400 M⁻¹s⁻¹ in a single design round. We calculated the cost per functional enzyme across four generations of computational catalysis: $1,106 in 2008, $338 today.
OpenAI Wants 900,000 Memory Wafers a Month. That’s 56% of Today’s Global Output. South Korea Just Bet $576 Billion It Can Deliver.
South Korea committed $576B to chip manufacturing because one customer wants more than half the world's DRAM. An original demand-absorption analysis shows even doubling capacity barely closes the gap.
421 Electrodes Just Outperformed 1,024. The Brain-Computer Interface Industry Is Optimizing the Wrong Variable.
Per-electrode efficiency across five BCI platforms reveals a 48-59× gap. The company with the fewest electrodes achieves the highest data rate. Electrode count is a marketing metric, not a performance metric.
An AI Law Firm Just Won Its First Trial. It Charged £400. The Losing Side's Lawyers Cost Thousands.
The Economist calls AI lawyering's results 'mixed.' The access-to-justice math says we're closer than the skeptics think.
90,000 Patients Are Waiting for a Kidney. A 69-Edit Pig Organ and a New Antibody Just Produced Zero Rejections After Six Months.
CRISPR pig kidneys and a precision immunosuppressant converge in one protocol. Phase 2 data: eGFR 74 vs 61 at 18 months, zero rejections after six months. The $8.9 billion waitlist problem has a new answer.
OpenAI Designed Jalapeño in 9 Months. Google Took 3 Years for TPU. The 20 Engineers Who Built Both Explain the Difference.
At $475M estimated development cost and $1.25B in projected annual savings, Jalapeño pays for itself in under 5 months. But every company fleeing Nvidia's monopoly is hiring the same arms dealer.
Hyundai Is Deploying 25,000 Robots in Its Factories. The Data They Generate Is Worth 1.5× More Than the Workers They Replace.
At $108,000 per robot per year in projected data revenue versus $70,000-$75,000 in labor savings, the economics of Hyundai's Atlas deployment flip the humanoid narrative: the real product isn't automation. It's training data for every other robot on Earth.
Everyone Quoted CATL’s $35/kWh Sodium Battery Price. The Number That Actually Matters Is 3.6× Better Than Lithium.
CATL unveiled the first field-validated sodium-ion grid storage system on June 22. Its 15,000-cycle lifespan creates a lifecycle cost gap that dwarfs the cell-price advantage everyone keeps citing: $2.59 vs $9.40 per megawatt-hour per cycle.
Microsoft’s Quantum Program Has a 33% Retraction Rate on Its Key Papers. It Just Promised a Working Machine by 2029.
Two retracted Nature papers, two editorial alerts, and a new formal critique make five high-profile challenges across roughly six major publications. No other quantum computing lab has retracted anything. Google's Willow chip already operates below the error-correction threshold. The cost-per-demonstrated-logical-qubit math is merciless.
Friedreich Ataxia Has One Approved Drug. It Costs $370,000 a Year and Doesn’t Touch the Heart Failure That Kills Most Patients.
A single injection of AAVrh.10hFXN gene therapy reduced cardiac damage markers in 15 of 17 patients. Skyclarys, the only FDA-approved treatment, runs $370,000 per year and addresses neurological symptoms only. The lifetime cost math says gene therapy breaks even in under six years.
More articles load as you scroll.