April 20, 2026
When your codebase won’t sleep
Tim Davis – Probabilistic engineering and the 24-7 employee
“Robot interns” pull all-nighters while devs sleep — comments are split between hype, dread, and LOLs
TLDR: An AI exec claims “robot interns” now crank code overnight, turning workers into managers of nonstop agents. Commenters are split between believers citing sci‑fi‑level workflows, skeptics warning of messes and energy‑fragility, and jokers asking if humans now write like AI—raising real questions about jobs, quality, and who cleans up the code.
Tim Davis says the new normal is probabilistic software and the 24‑7 employee—not a person working all day, but a human backed by tireless AI “agents” pushing code while you sleep. His side project allegedly spins up overnight pull requests like a factory line. The crowd? Oh, they had feelings.
Skeptics rolled in first. One poster wondered where the real wins are, saying they “haven’t seen much real output” and prefer a guided dev style with humans steering. Another fired a doom flare: if energy markets wobble or global chaos hits, what happens to your “robot interns”? Meanwhile, a salty veteran grumbled the machines (and their human copilots) are just multiplying bad code, warning: “Good luck to anyone cleaning up the mess.”
But the believers showed up, too. One insider swore small teams are already doing this with tight human-in-the-loop loops, stacking abstractions and letting models review changes—calling it “science fiction becoming aware.” And then the existential crowd joined the brawl: “I feel like the internet is programming me,” one user joked, asking whether AI writes like people or people write like AI.
Davis’s claim that roles are splitting—super‑operators rising, “agent babysitters” stuck in career purgatory—lit a nerve. It’s a soap opera: hype vs. cleanup crew, pragmatists vs. prophets, all arguing over who’s the boss when the bots never clock out.
Key Points
- •The author argues software development is becoming probabilistic in AI-native teams, weakening traditional deterministic assurances.
- •He built a system called Compound Loop to orchestrate multiple models that autonomously write, review, and merge code overnight.
- •This workflow creates a “24-7 employee” model via parallelized agents, changing how work is scheduled and triaged.
- •Roles are diverging: some operators move up to product and architectural levels, while others shift into fragmented tasks like spec writing and prompt-based reviewing.
- •The author asserts fragmented roles will be less valued and paid, widening pay gaps; scarce work is shifting toward AI infrastructure.