Forgot your password?
typodupeerror

Comment Re:A good trade off (Score 1) 65

It is an actually good use of LLMs this time around.
You get a LLM, tell it to look for exploits and bugs, and it will spit you 5000 answers, then you use another one or something less "prone to errors" to check if these answers are real.
50 will be and brand new, and then a human goes there and actually fix these 50 bugs.
The trick with LLMs is to assume they will be wrong and engineer around it, so this example above and many others are good use case, vibe code (as in someone that don't know programming using it to generate code) is the worst use case possible.

Comment Re:Relavant XKCD comic (Score 1) 80

Distros are more of "the taste of the consumer", specially in basic appearance and utilities.
As long you don't make an incompatibility trainwreck, you can make a bunch of different distros to try to get to different tastes.
The fragmentation is only a problem if you get stuff like "this distros can only run X programs".
If all of em can run all of the software, it's fine and even desirable.

Submission + - WSJ: Corporate America Has Suddenly Decided to Stop Blowing Money on AI

theodp writes: "Companies across the country are coming around to a radical idea with the potential to upend the industry powering the global economy," reports the WSJ. "They don’t have to blow their budgets on AI. Fed up with ballooning costs, companies big and small are starting to use lower-priced models, including some built in China. In many cases, they are adding the new, cheaper models alongside OpenAI and Anthropic’s products, shopping a la carte for their artificial intelligence."

"Being economical—or tokenomical—is a dramatic reversal in mindset. Just a few months ago, it was a badge of honor to be using AI so much that you spent a lot on tokens. Companies rewarded employees for tokenmaxxing, flashing leaderboards that showed who had spent the most. Now they are thrift-maxxing."

"Strategies to lower AI costs include limiting access to top models for new hires and using the most advanced AI systems to plan how tasks will be completed before turning to cheaper models for the execution, Cursor’s [Mike] Saeks says. Cursor recently ran an experiment to evaluate the cost of building a web browser from scratch. Doing the entire task on OpenAI’s GPT-5.5 cost a little more than $10,000. Using Cursor’s Composer coding model in combination with Anthropic’s Opus 4.8, cost $1,339."

Correction: The costs cited by the WSJ were evidently to recreate SQLite functionality, not to build a web browser from scratch, an earlier experiment that Cursor explained involved running coding agents autonomously for weeks and burning trillions of tokens to build software that Cursor's CEO boasted "*kind of* works", prompting The Register to credit Cursor for "proving AI can write shoddy code at scale".

Comment Right once, patch away (Score 2) 77

To find a security flaw, you only need to be right once.
So if you have a machine that has a shitload of false positives, and have a way to filter em quickly, you end up with a bunch of true positives.
Now to code, you ideally want to always be right, which is not quite ideal for a machine that does a lot of false positives.
It's a pretty fun scenario, specially if you're not the only one running the security flaw finding machine.

Slashdot Top Deals

Logic is a systematic method of coming to the wrong conclusion with confidence.

Working...