victorlegros.bsky.social
@victorlegros.bsky.social
Looking up feature flag implementations for TypeScript (open to recommendations BTW), and came across this after pursuing a line of investigation: dougseven.com/2014/04/17/k...
Knightmare: A DevOps Cautionary Tale
In the first 45-minutes the market was open the faulty software deployment sent millions of child orders into the market resulting in 4 million transactions against 154 stocks for more than 397 mil…
dougseven.com
November 12, 2025 at 4:09 PM
Piecing together a hardwood floor repair is a Knapsack problem. And thus NP hard. Really glad to be using my comp sci masters in practical affairs 😂.
November 2, 2025 at 5:53 PM
Maybe the solution to AI enthusiasm is hedonic treadmill exhaustion? www.newyorker.com/news/fault-l...
Sora 2 and the Limits of Digital Narcissism
What we enjoy about generative A.I. may also be its ultimate limitation: we want to see ourselves.
www.newyorker.com
October 26, 2025 at 2:20 PM
Yay! Cool algorithm to dive into about maintaining robot balance: arstechnica.com/ai/2025/10/w...
Why iRobot’s founder won’t go within 10 feet of today’s walking robots
Rodney Brooks says humanoid robots pose hidden safety challenges and won’t learn dexterity from video alone.
arstechnica.com
October 3, 2025 at 3:15 PM
The path that led me to live, work, and eventually become a citizen in the US is being severely curtailed. Gotta prep those mental gymnastics to justify how this will help the US be a destination for talent and innovation. www.bbc.com/news/article...
H-1B: Trump adds $100,000 fee for skilled worker visa applicants
"Train Americans," the US commerce secretary told big companies, some of which have relied on foreign high-skilled workers for years.
www.bbc.com
September 20, 2025 at 6:11 PM
Yup, folks would be lining up to buy Chrome. User acquisition is so pricey, and Chrome is prime realestate… for now. arstechnica.com/gadgets/2025...
Perplexity offers more than twice its total valuation to buy Chrome from Google
Google may soon be ordered to sell Chrome, and Perplexity is ready.
arstechnica.com
August 14, 2025 at 12:47 AM
Here’s a cogent argument for not dying in a blaze of glory, c/o article comments: arstechnica.com/ai/2025/07/e...
Exhausted man defeats AI model in world coding championship
“Humanity has prevailed (for now!),” writes winner after 10-hour coding marathon against OpenAI.
arstechnica.com
July 19, 2025 at 9:15 AM
Reposted
Thank you for EVERYTHING, @pbs.org ✨ We love you, we will never give up on you!The support from viewers like us for you is eternal! Public media teaches us kindness, teaches us to enjoy learning, teaches us that we all belong. PBS is my reason to stay. PBS, always and forever ❤️‍🩹💙🩵💚
July 19, 2025 at 12:44 AM
Read this article, and much of it rings true with my experiences, just dialed up even further. Definite ‘superstar economy’ at play in the market. Meanwhile the industry is plagued by ingrained habits and an unwillingness to confront difficult truths. arstechnica.com/gaming/2025/...
What’s wrong with AAA games? The development of the next Battlefield has answers.
EA insiders describe stress and setbacks in a project that’s too big to fail.
arstechnica.com
July 16, 2025 at 5:45 PM
Something similar has come up multiple times in my career: how do you measure success, and can the metric also be the target? The moral hazard is also sometimes intentional, especially with big money on the line. arstechnica.com/ai/2025/07/a...
What is AGI? Nobody agrees, and it’s tearing Microsoft and OpenAI apart.
Several definitions make measuring “human-level” AI an exercise in moving goalposts.
arstechnica.com
July 14, 2025 at 3:17 PM
Really insightful article comment. That high consequence scenarios require expertise, and why “good enough” isn’t . arstechnica.com/ai/2025/07/a...
AI therapy bots fuel delusions and give dangerous advice, Stanford study finds
Popular chatbots serve as poor replacements for human therapists, but study authors call for nuance.
arstechnica.com
July 13, 2025 at 2:02 PM
I guess AI help is mid?: www.reuters.com/business/ai-... AI slows down some experienced software developers, study finds | Reuters
AI slows down some experienced software developers, study finds
Contrary to popular belief, using cutting-edge artificial intelligence tools slowed down experienced software developers when they were working in codebases familiar to them, rather than supercharging their work, a new study found.
www.reuters.com
July 12, 2025 at 10:53 AM
A sensationalist title, though interesting about the subtle assumptions about trust and anything self-referential. Also, given that information diffuses slowly and people don’t like grunt work like auditing, means it’s bound to happen. www.quantamagazine.org/computer-sci...
Computer Scientists Figure Out How To Prove Lies | Quanta Magazine
An attack on a fundamental proof technique reveals a glaring security issue for blockchains and other digital encryption schemes.
www.quantamagazine.org
July 9, 2025 at 4:57 PM
I mean, I thought his focus and commitment was impressive back in 2021. And he’s certainly throwing his weight now. Big question is, is that enough to succeed?

(Future narrator: it was not)

arstechnica.com/ai/2025/07/m...
Meta’s “AI superintelligence” effort sounds just like its failed “metaverse”
Zuckerberg and company talked up another supposed tech revolution four short years ago.
arstechnica.com
July 3, 2025 at 8:57 PM
Just finished @adambecker.bsky.social “More Everything Forever”. A well researched summary of the ideas pervading Silicon Valley, and calls out the limits of those, plus historical antecedents. Not a recipe for becoming wealthy, but for maybe seeing through some questionable arguments.
June 29, 2025 at 6:16 PM
Years late to the party, but finally doing some Kerbal Space Program!
June 26, 2025 at 9:50 PM
I think about this type of failure a lot. It's why I'm a product manager; to help teams cross that 'valley', keep money flowing and customers invested.

www.bbc.com/news/article...
What happens when a high-tech project fails?
What happens when a high-tech project fails?
Workers at Reaction Engines felt they were close to completing a revolutionary jet engine.
www.bbc.com
June 3, 2025 at 4:32 PM
I guess I can always fall back to my MS-DOS through Windows XP expertise to put food on the table => www.bbc.com/future/artic...
Still booting after all these years: The people stuck using ancient Windows computers
As technology marches on, some people get trapped using decades-old software and devices. Here's a look inside the strange, stubborn world of obsolete Windows machines.
www.bbc.com
May 29, 2025 at 1:44 AM
Incredible use of ten minutes. Once again reminded about where we are with AI and lessons from history => podcasts.apple.com/us/podcast/5...
Spreadsheet
Podcast Episode · 50 Things That Made the Modern Economy · 05/13/2019 · 11m
podcasts.apple.com
May 28, 2025 at 8:40 AM
Wrote a post on LinkedIn about a missed opportunity for Duolingo => www.linkedin.com/pulse/why-i-...
Why I ‘prestiged’ in Duolingo, and why you should too
Gamification. The idea that you can turn a task or chore into something fun, by layering on flashy elements and positive reinforcement.
www.linkedin.com
May 23, 2025 at 6:07 PM
“First neural nets, now evolutionary algorithms... All the old machine learning techniques are new again!”

From the comments of =>

arstechnica.com/ai/2025/05/g...
Google DeepMind creates super-advanced AI that can invent new algorithms
AlphaEvolve has already made Google’s data centers more efficient and improved Tensor chips.
arstechnica.com
May 15, 2025 at 2:17 PM
Read weeks ago, but still an incredible article =>

arstechnica.com/space/2025/0...
Starliner’s flight to the space station was far wilder than most of us thought
“Hey, this is a very precarious situation we’re in.”…
arstechnica.com
May 11, 2025 at 2:18 AM
Surprised by the line, "may not lead to any immediate applications" in the article. Maybe I need to read the underlying paper. www.quantamagazine.org/undergraduat...
Undergraduate Upends a 40-Year-Old Data Science Conjecture | Quanta Magazine
A young computer scientist and two colleagues show that searches within data structures called hash tables can be much faster than previously deemed possible.
www.quantamagazine.org
February 10, 2025 at 8:23 PM
bbc.com/future/artic...

"Embodied AI"; seems like some amount of pre-training and feedback control could enable comparable dexterity.
Human hands are astonishing tools. Here's why robots are struggling to match them
Our hands perform thousands of complex tasks every day – can artificial intelligence help robots match these extraordinary human appendages?
bbc.com
January 27, 2025 at 10:51 PM