cephei

There are so many use cases for supremely fast offline models. The first thing that comes to my mind is for real-time video processing or other non-textual content in real time.

In a year or more, the audience will likely be an agent instead of a person. It'll be interesting to see how that shifts language and article formats.

Many points about maintainability that this article makes seem to hold, especially with update and version output changes, but it doesn't even offer the simplest alternative. Most of these supported commands have flags to strip out noise and reduce output. Maybe agents aren't well trained on these.

As a side note, has anyone tried a dual agent setup where the command output is proxied through a lightweight local model? I can imagine a scenario where all tool output is filtered through Qwen or similar locally to compact the tool output.

As I understand it, they modify the image by applying a special Gaussian noise filter which affects each pixel in the image in subtle (possibly not reversible) ways. The detecting service will look for this noise pattern to flag it, so even a part of the image is enough to know it was generated by AI.

In my experience, I've seen engineers try to take on more work to get promoted, but the key issue is that they were doing more work at their own level instead of focusing on work that would be their responsibility if they promoted. If an IC takes on more and more IC work instead of management responsibilities, it's harder to promote them.

I think primarily this victimizes all those all ready victimized by the CSAM in the training material and also generally offends the collective sense of morality our society has.

It certainty depends on how the network is doing the packet scanning and the other measures in place. If the security measures include a allow list of DNS servers, then there's little that can be done to bypass it.

It's hard to know exactly how much energy this is pumping out and how focused the beam is, but suffice to say anyone in the way might start to boil.

This may be more palatable for the company rather than calling the extra money for being in the office a "commute stipend", even though they are effectively the same.

It's fairly fundamental in languages, just hidden behind syntactic sugar like if/else and try/catch/finally. Outside of those it can be useful for de-duplicating cleanup and exit code from a function.

Even with the 1 in 10000 false negative rate, I bet someone is doing the cost calculation of risk vs how many hours it would take for a doctor to check 10000 scans. Doctors themselves are not perfect so they may even have a higher error rate.

I believe the relevance comes from the fact that both are built on Electron.

The closest parallel I've found for this in English is the phrase "active listening" which includes this kind of interjection to reassure the speaker that they are heard.

I have seen this work several times. In many cases, it's easier to change roles inside a company than to switch to a dev role at a new place when you have no developer experience. Starting out in QA is a great way to develop domain knowledge, which can be more important than programming skill in many businesses.

The whole complex comes in at about 34 km^2 which puts the population density at about 10 times the density of Manhattan. The water and energy needs alone will require several key innovations before this becomes even a pipe dream. Household and public space water usage could be a billion gallons of fresh water a day.

If there is sensitive data to remove and the history is important to keep, then GitHub has some recommendations for scrubbing the history

https://docs.github.com/en/authentication/keeping-your-accou...

Does HN have any Facebook integrations that could have caused the slowness seen earlier? Or it could be that some part of the network infrastructure is shared between the two and was overloaded.