So Mythos actually benefits software security, since developers uncover vulnerabilities that would otherwise go undisclosed?

Would mutation testing help in this case? I see a strong case for a test that wasn’t actually testing the right subject. Maybe it wouldn’t help for more complicated cases??

Is building power tool UIs their moat? I expected something different, but their videos don’t show much. A lot of the same workflows seem possible with Jira cli, gh/glab cli, and CC. Maybe Im not understanding their offerings.

I’ve had a similar experience this morning. I tried copilot Claude and it seems to follow instructions correctly but Claude code with opus seems to be worse than Gemini 2.5 today.

How many days until we get hacked? I’m thinking definitely within the next month since they have about 30+ engineers running with yolo permissions and are specifically told to not review code. I wonder how many companies are doing this besides mine.

lol the VP and CTO at my company asked us to do this not just for claude but for codex, gemini and opencode.

Is this similar to www.boris-belousov.net/2018/11/12/c... and math.mit.edu/~apost/cours...?

As someone who was a recent grad only a few years ago, would this apply to them too? Meaning junior engineers should spend less time typing out code and more time exploring design spaces with Claude.

The article is mostly telling academics to use ai. I don’t think it ever gets into the author’s expertise to the point it’s not readable for others. The article makes the worst arguments for using AI, which is no surprise given that it was written with almost zero effort.

you have nothing to add bc it's a good article or bc it's a terrible article?

Psych joke?

Doesn't page 10 have 5 citations?

It's absolute cinema *insert Scorsese meme*. Honestly, worried for David, he's getting ganged up on in there.

Does it matter (genuine question)? It's really hard to say claude just made this up from nothing right? Zire is also in the baby names website as a potential baby name. If we believe LLMs just strings together words (tokens) from the training data then this example fits that belief.

"Let us all applaud the work done by Forest Department and volunteer Father and son duo Tanmaysingh Zire and his father for rescuing a tiger cub safely." - 2020 from a facebook post.

I didn't know you were joking. I wasn't being ironic. I am genuinely trying to figure out what you meant by "zire" not being in the training dataset.

What was the prompt to claude? How is "zire" not in the training corpus? "Zire" is definitely associated with Tiger cubs in the training data.

I listened to a few of his podcast episodes and I’ll never get that precious time back. He’s just riding the wave of outrage, exploiting people’s fears about AI taking their jobs. People need a light at the end of the tunnel, and he’s filling it with lies, hate, and false expectations.

These people aren’t engineers.This is super irresponsible. “watching it stream by” don’t overburden yourself.

Thanks for shaming this type of LLM usage. It’s important to make the distinction between this and the rigor you bring when using them.

workaround section of https://github.com/anthropics/claude-code/issues/16470

Maybe you've already seen the github issues but in case you haven't tried this yet. github.com/anthropics/c...

how often does this happen?

I can't believe these people are taken seriously by the community.
@ahl.bsky.social Web3 is so back!! 2025 1 year prediction.

"...a "tech geek" begins to thoroughly analyze a Web3 product, the market briefly realizes that BagsApp might be more than just a Meme frenzy-it could be a fundamental reconfiguration of the creator economy. Many believe he embodies the most hardcore and pragmatic engineering mindset from SV"

"Steve Yegge received a message on LinkedIn saying that someone was sending him money (initially $49,000), and mentioned that developer Geoffrey Huntley had just received $56,000 the previous week. His wife said it was a scam, and his intuition told him it was a trick, but he still tried it."

How do people keep falling for this crap over and over and over? Quote from www.kucoin.com/news/flash/b... :

Damn I don’t understand why anyone would go with ChatGPT over Claude.

Dm-ed you the link to the position.

most*

A lot of what you wrote stood out to me. The part that stood out the post is:
"I have a penchant for driving investigations to the ground and considering interactions of complex systems as a whole."

My company is hiring for Staff Embedded Software Engineer. If you are interested, I can refer you.

I would agree with your take if I were working at Oxide (because y'all bring rigor to a complicated system). However, not every company has the expertise and values that Oxide employees have. When you have people letting agents loose on a production system, it's hard not to be cautious. I'm biased.😔