πΊ Deep Dive 3 β Nathan Sobo (Zed Founder):
Why IDEs won't die β even in the age of AI coding.
Terminal-based tools like Claude Code are mindblowing for making progress, but terrible for reviewing what the AI did. You only see a 10-line excerpt.
"If you believe the IDE is going to die, that requires you to believe humans won't need to interact with source code anymore."
πΊ Deep Dive 2 β Sanjit Biswas (Samsara CEO):
The next AI revolution happens while you sleep.
The "third shift" (midnight to 8AM) is currently unworked by humans. Physical AI can run logistics, monitoring, and maintenance during that time β adding 8 hours of operational capacity to every day.
Physical AI runs on 2-10 watt edge devices, not cloud servers.
Jim Chanos on why data centers are a bad business:
The old-timers run mid-to-low single digit pre-tax returns on capital. The new builders? Many are recycled Bitcoin miners.
Neo-clouds like CoreWeave are just equipment leasing β buy a chip from Nvidia, lease it to a hyperscaler. Middlemen trading at higher multiples than the companies that control their supply.
"You're paying for that now."
Jim Chanos dismantles "data centers in space":
The physics don't work. Launch costs aren't low enough. Radiated heat requires enormous radiators. Redundancy is a nightmare β you can't just send a tech with spare parts.
And the one rocket all of this depends on β Starship β hasn't even reached orbit in 12 flights.
"A lot of this is really out there a number of years. The problem is you're paying for that now."
Jim Chanos breaks down the SpaceX IPO:
$75B raise. ~$2T valuation. $19B revenue. Negative free cash flow.
"This is really a hopes and dreams IPO."
The entire valuation rests on promises β Mars colonies, moon factories, data centers in space. Not on the actual business.
Chanos: "In bull markets you put a premium on promises. In bear markets you put a discount on reality. We're clearly in the former."
The takeaway isn't despair. It's that the fight is never over, and the tools exist. Use Signal. Support the EFF. Pay attention to the Earn It Act and similar bills.
The people who built the encryption we rely on had to sue the government to do it. The least we can do is use what they fought for.
Full episode: https://t.co/1YNV5K6yzZ
Follow @entradox for more signal like this.
π§΅ 7/7
Everyone's scared of AI surveillance for the wrong reason. They think it's about more cameras. More facial recognition. More tracking.
The real danger is simpler and worse: AI makes surveillance permanent. Every photo, every recording, every data point becomes searchable forever.
π§΅ 1/7
Cohn's conclusion is sobering: privacy is like free speech. You don't "solve" it. You defend it continuously. The moment you stop paying attention, it erodes.
The AI multiplier makes this harder. It doesn't create new surveillance capabilities β it makes existing ones permanent, searchable, and actionable at scale.
π§΅ 6/7
Here's the wildest part: Meta is now arguing in court that encryption is a "product defect." A legal argument that flips the entire Bernstein victory on its head.
The same company that built the world's largest surveillance machine is now saying privacy features are a bug. That's not irony β that's a coordinated legal strategy.
π§΅ 5/7
The encryption fight is structurally endless. Law enforcement's desire to eliminate private conversations never goes away. Neither does our need for them.
Every generation has to re-fight this battle. The 90s gave us the legal foundation for encryption. The UK, Australia, and Canada are now actively trying to tear it down.
π§΅ 4/7
Think about what "permanent and searchable" means in practice. A photo taken today of you at a protest. A video of you outside a building. Right now, it's noise in a sea of data.
With AI, that noise becomes a searchable database. One political shift, one grudge, one investigation β and that noise is evidence.
π§΅ 3/7
Cindy Cohn ran the EFF for 11 years. She fought the Bernstein encryption case in the 90s β the one that made Signal and HTTPS possible. She's watched this movie before.
Her take: the creep factor isn't the hardware. It's the AI layer that turns a police robot dog into an indelible record that can be weaponized against anyone, anytime, retroactively.
π§΅ 2/7
"The center of the universe has shifted from finance to semiconductors." AI isn't a sector. It's the substrate everything else is building on. π§΅ 1/7