Nature published a peer reviewed paper arguing that Microsoft's claims of a quantum computing breakthrough last year are based on bugs in their Python code and selectively choosing which data to base their results on.

The author of the paper claims "I demonstrate that Microsoft's tune-up software is flawed and that coding errors resulted in incorrect statements to peer reviewers"

This is extremely embarrassing for Microsoft and they've pushed back on the paper's claims.

https://www.theregister.com/research/2026/06/24/boffin-claims-microsofts-supposed-quantum-leap-does-not-compute-due-to-basic-python-errors/5260489

The reflecting pool is a metaphor for everything that’s wrong in America today.

Arrogant people with power who don’t understand how anything works, making decisions that will predictably fail and we have to deal with the ugly aftermath.

Every day feels like we’re unwilling participants in a cautionary tale that future generations will study in disbelief.

The #1 reason I vibe coded a replacement for the Google Health app for my Fitbit data is that I’m a biphasic sleeper.

I wake up in the middle of the night, read and post for a few hours then go back to bed.

Google Health only tracks my first sleep as my nightly sleep. My app combines both. 😴 🛌

Elon Musk, CEO of SpaceX, posted on X on Sunday that the company “might be able to reach approximately” $1 trillion revenue in 2030.

If this happens, I will fry up my nicest pair of shoes and eat them on Instagram live. Dude will say anything to pump the stock price.

https://www.cnbc.com/2026/06/15/spacex-stock-record-ipo-debut.html

When it rains, it pours. Anthropic is facing a class action lawsuit over the claim that its “Max 5x” ($100 a month) and “Max 20x” ($200 a month) plans deliver far less usage than advertised, and the court has been asked to make the company refund subscribers.

Welcome to big tech, Dario.

https://www.wsj.com/tech/ai/anthropic-sued-over-limits-on-its-200-a-month-ai-plans-e2a109e4

I hear a lot about AI-driven layoffs and it’s easy to assume every white-collar job is affected. So far the strongest examples are in tech and banking.

My guess is hiring eventually rebalances. Big tech shrinks while smaller companies hire more as AI increases the value generated per developer.

A major reason I built my replacement for the Google Health app is that Google replaced most of the dashboard on the Home Screen with an LLM coach that provides proactive health tips.

Like all LLM-based features, it’s often wrong. This is a great example of how shoving AI into an app makes it worse.

They should invent the civilian version of war crimes for Elon Musk.

https://www.currentaffairs.org/news/how-elon-musk-killed-hundreds-of-thousands-of-people

In 1999, the PowerMac G4 was was classified as a "restricted supercomputer" by the U.S. government due to its processing power breaking the 1-gigaflop threshold. It was then banned from being exported to 50 countries.

So Apple ran ads bragging about the fact.

History doesn’t repeat but it rhymes.

For the first time in years I have multiple startup ideas.

The common thread is that they are bets against two assumptions:

1. AGI arrives soon and fixes LLM reliability.

2. AI usage stays cheap enough that companies never need to manage it.

Startup idea #1: AI slop protection for businesses.

Software has linters, type checkers and tests. AI-generated work mostly has vibes.

You can’t stop people from generating AI slop. It’s too tempting. So the business is catching it before it reaches customers, courts, regulators or production.

Startup idea #2: AI ROI measurement.

The question isn’t whether tokens get cheaper. Cloud got cheaper per unit while AWS bills exploded.

AI will be similar. Once usage is big enough to matter, companies will need to know who is creating value and who is just burning tokens.

I may finally take my sabbatical and spend time exploring if there are good products I can implement around these ideas.

After all, ideas are cheap and execution is what matters

AI is a force multiplier for workers but it could make one person 5x more productive and the other 0.5x as they generate AI slop that their coworkers need to deal with.

The challenge will be how for test for this in interviews and reflect in performance evaluations going forward as people’s impact diverges.