AI Just Designed a Drug and Struck a Bomber Fleet

  | Submit tool | Advertise | Tools Published Today |

Wednesday Deep Dive

_{(Reading Time: 4 minutes)}

The Wednesday Deep Dive takes a detailed look at what's new in AI. Each week, we share in-depth insights on new tools, proven prompts, and significant developments - helping tech professionals work smarter and stay ahead.

This week, we spotlight two major AI breakthroughs—one at the molecular level and the other on the battlefield. Together, they show how open infrastructure, hardware, and synthetic intelligence are upending both the lab and the war room.

🧬 AI-designed drug shows results for lung disease

🚁 Ukraine uses open-source AI to combat Russia

Let's dive in.

🌐 AI News

🧪 AI Hits Clinical Milestone With First-of-Its-Kind Drug for Lung Disease

A drug designed end-to-end by AI just passed a mid-stage clinical trial, and it's aiming at one of the toughest diseases in respiratory medicine.

Biotech firm Insilico Medicine, headquartered in both Boston and Hong Kong, announced that its lead drug candidate, rentosertib, improved lung function in patients with idiopathic pulmonary fibrosis (IPF), a fatal condition marked by progressive lung scarring. Published in Nature Medicine, the findings suggest this is the first time a fully AI-conceived drug, based on an AI-generated disease hypothesis, has demonstrated clinical efficacy in human trials.

🧬 What’s New:

• Rentosertib cut levels of a protein that AI identified as disease-linked.

• At the highest dose, patients saw statistically significant lung improvements.

• AI designed both the mechanism and the molecule, marking a clinical first.

The study included 71 patients across China and ran for 12 weeks. While not large, the trial showed that the drug worked exactly as predicted, down to its impact on specific molecular targets.

📊 Results Breakdown:

✅ Lung function improved in the high-dose cohort.

🧪 Biomarker analysis confirmed molecular targeting.

⚠️ Some patients experienced liver side effects.

❌ A small subset worsened during treatment.

CEO Alex Zhavoronkov called it "one of the best results people have ever seen" for IPF. He added that Insilico’s approach enabled them to reach trials in just 18 months. That’s a third of the traditional timeline.

👀 Strategic Implications:

📈 Insilico has >30 AI-driven drug programs. This trial serves as a live demo for investors and pharma partners.

💸 The company says it has the funding for larger trials but is now soliciting regulatory and industry feedback on its path to approval.

🌏 Trials are expanding in the U.S., with global regulatory timelines expected to span at least two more years.

🔍 Why It Matters:

• AI Drug Design Just Got Real: This isn’t just AI-assisted discovery. It’s AI from end-to-end—disease theory to compound to clinical data.

• Time-to-Trial Crushed: 18 months to IND readiness could transform R&D cost models.

• Proof of Platform: If validated in future studies, Insilico’s stack becomes a roadmap for biotech 2.0.

⚠️ Caveats & Considerations:

• Small sample size and short trial window.

• Liver toxicity must be closely monitored.

• Efficacy still needs replication in larger, longer studies before any regulatory approval.

Even so, the implications are massive. If rentosertib clears larger trials, it’ll be a flag planted in the ground for AI-led medicine.

🌐 AI News

🔥 ArduPilot and DIY Drones Wipe Out a Third of Russia’s Strategic Bombers

In one of the most audacious military strikes in recent history, Ukraine used open-source autopilot software and consumer-grade drones to destroy 34% of Russia’s strategic bomber fleet in a single day.

The operation, dubbed “Spider Web,” hit three air bases hundreds of miles deep in Russian territory: Belaya, Olenya, and Ivanovo. The drones were powered by ArduPilot, a 100% free, open-source UAV control stack originally built in a California garage using LEGO parts and Arduino boards.

💥 What Happened:

• 117 drones launched from concealed sheds and truck containers.

• Operated by small quadcopter teams across three time zones.

• Targeted bombers on open tarmacs with known GPS coordinates.

• Execution: flawless. Russian Telegram channels posted videos confirming the hits.

The raid’s preparation took 18 months and was run by Ukraine’s Security Service (SBU). Most drones didn’t even rely on Starlink. They used basic modems and Raspberry Pi-style boards to connect via Russian mobile networks, a workaround that allowed Ukraine to dodge GPS jamming and GLONASS interference.

🧠 Why ArduPilot?

• Fully autonomous and GPS-capable

• Loiter and stabilization modes allow for signal-loss fallback

• Free and modifiable, with no licensing limits

• Built-in failsafes are ideal for complex, multi-drone ops

🛠️ Background on the Software:

• Started in 2007 by Chris Anderson (former WIRED EIC), Jordi Muñoz, and Jason Short.

• Originally made for hobbyists via DIYDrones.com

• Has since powered UAVs for agriculture, ocean research, cinematography, and now, apparently, military-grade sabotage

📉 Cost Asymmetry:

• Ukrainian DIY quadcopters: Hundreds of dollars each

• Russian bombers destroyed: Tens of millions in value

• This is asymmetric warfare at its most brutal and efficient

📣 Open Source Ethics in the Spotlight:

ArduPilot’s devs didn’t publicly comment, but one posted anonymously on Reddit:

“We won't discuss or debate our stance, we [focus] on giving you the best tools to move your [vehicles] safely. That is our mission. The rest is for UN or any organisms that can deal with ethical questions.”

ArduPilot’s code of conduct discourages weaponization, but as open-source software, there’s no enforceable license preventing wartime use.

🎯 Strategic Fallout:

• Exposes vulnerabilities in Russia’s defense posture, planes left in open rows, visible via satellite

• Demonstrates the rise of low-cost, high-impact drone warfare with no defense playbook

• Showcases open-source software as a national defense tool, whether its creators intend it or not

📦 Tech Notes:

• Russia’s larger “Baba Yaga” drones were previously captured running on Starlink + ArduPilot

• The Spider Web drones were much smaller and cheaper, but equally lethal

• Analysts believe latency was mitigated using ArduPilot’s autonomy and GPS backup features

🧠 Expert Insight:

Kelsey Atherton, editor at the Center for International Policy, believes that the biggest innovation here is resilience by redundancy. Cheap, disposable drones with good-enough software outperform bespoke defense systems 10x their cost.

Atherton also pointed out that the only real fix to this kind of attack might be as low-tech as it gets: “The biggest fix would have been hangers with doors that close. It's an intelligence failure and a parking failure.”

Want to Unsubscribe from future emails? Just hit “unsubscribe” below.