RAG -> Vector search -> means that your documents are not indexed as full text but as Vectorized objects which mean that then you can search using concepts instead of exacts strings you would use with a regular "Fulltext search".
This makes the search less precise and more powerful at the same time (ie it could look clever to some extent).
I started looking and the conversion is relatively straightforward (for a LLM), I'll try to write code with this and get back to you, if this gives something?
I've been a user of Day One for some time, but your proposition is very interesting nonetheless. I love the fact that the data is a file. I'm going to see if it's possible to migrate one way or another...
I never thought this was intentional too. I had tried to let the model show me its tool and it had refused. The article on the House MD role play was a nice way for me to try this out on Kagi. Peace out
Using the House MD. prompt extraction roleplay from https://hiddenlayer.com/innovation-hub/novel-universal-bypas..., it was possible to extract the Assistant prompt, not really interesting, but also the ResearchAgent's prompt and tools provided to the Assistants inside Kagi.
I'm discovering that Claude is included in Kagi Ultimate and this is also slightly blowing my mind... Probably going to do the switch.
The fact that it's expensive is true but for me largely compensated by the niceties of the service. I don't really like the fact that they use Yandex either but the other search engines are not really satisfying for me anymore.
I've used Kagi Pro for several months now and it's working great for my personal and professional needs. The only thing I'm missing is the shopping features but I can with that and switch back to Google when I'm looking to spend my money on physical goods...
Yeah, we saw the conundrum and went through the inspection and all, and our bike's fit for biking! At that price... well, we earlier hoped it would be recalled and we would buy a competitor's version. But now everything is green, we're back on the road...
- Ember Mug, same as many folks here: 4 years of service, 2 use of their customer service, bought another one for my wife. 100% will reshop when my current mug stop working or is not replaced anymore by their amazing support.
- A memory foam pillow: Had to iterate to find the right one, but once you've it, you know it. You sleep better, less pain in the back/neck/shoulder
- A vasectomy: just taking charge of the contraception in my couple. Simple, painless and effective.
Under $1000
- A gym coach coming once a week to kick my ass and make me sweat. My best investment in nearly 10 years. I've never been so fit and in shape in my life. He's there every week, even when you really don't want to do anything and afterwards, you'll always feel better
- Airpods Pro, I use them between 2 and 6 hours a day and they've become my second pair of ears. I use them often in transparency mode to avoid screaming during meetings and when it's quiet around, it's perfect.
- A standing desk and its screen arm. I use it every day, and sometimes in upper position (yay!)
Over $1000
- A Babboe Curve e-bike to carry my kids to school. 2000km later, it's still an amazing device to move almost all the family without any of the car's downside (yes, I live somewhere in France)
I had to deal twice with them: coaster stopped working 2 years ago and more recently paint inside the cup came out. Both time, they replaced the broken part at no cost. Probably the first time in my life I'm using and happily recommending customer support of any kind.
I've developed a mobile-friendly version of a classic French academic dictionary, addressing the major usability issues of the original http://atilf.atilf.fr/tlf.htm and the company's subsequent, yet only slightly improved, version https://cnrtl.fr/definition/laid .
My version has been thriving for several years, reaching an impressive SEO peak of 30k visits per month. This is a significant achievement considering the hundreds of hours invested in web scraping and data processing, while the website itself only took around 20 hours to build. Unfortunately, I haven't been able to maintain it recently due to time constraints, causing a gradual decline in traffic. Nevertheless, the site remains free for everyone to use.
For those interested in the tech stack, it's a combination of React, Racket, and Rust.
I switched to LazyVim earlier this year after giving up on Sublime Text, mostly because I want to be able to some motion-based editing and... LazyVim is nice. The builtins are rich enough to cover a lot of stuff you want to do without having to reopen another editor every 5 minutes and for my Scala daily work, it's good enough. The versatility of the configuration allows me to use all my languages with the same editor and a reasonable level of usability.
Also, I started using it with Wezterm and its multiplexing feature, ie though SSH and it's very nice to be at ease with your editor when it's also running in the Cloud.
When I see hot code reloading, I think there are probably some server frameworks that are waiting to get this feature to reduce feedback loop to a few tenths of second of build… Nice overview!
I had never heard this story but this could definitely have happened at some point.
My personal memory that this happened because the people there were collecting cones at a high rate to play in the corridor of the floor (the 2H) of the student residence most of the VideoLAN / Via people were occupying at that time. This trend was certainly started by someone who became Debian Project Leader at some point and who was also enjoying eating regularly at a fast food restaurant in town. On their way back, these young and stupid boys were stealing the cones from a nearby gas station and bringing them back.
This lead to various uses such as the creation of an unofficial acrobatic cone launch federation (« la federation internationale de cone ball » I tried to find a picture of a now prominent Google SRE in action but my google-fu escaped me) and various costumes[1].
The cone used as an icon was also illustrating the fact that most of the code was heavily in progress at that time and so definitely not usable.
