HN user
Didn’t read the full article but it starts with
The jet was perhaps the pinnacle of American engineering excellence. Its retirement signals an end to an era of American culture—and ambition.
End of American ambition? SpaceX landing is rockets… today! That’s apples to apples also, both aerospace. In other fields we have literally taught computers how to talk.
I only read Meditations. So yes, if he is not representative of the philosophy I have bastardized it.
Although, I am convinced Marcus Aurelius used Stoicism as a tool to control his emotions. Meditations reminded me of cognitive behavioral therapy. Repetitive soothing phrases to induce equanimity.
Stoicism in Ancient Rome was COMPLETELY about controlling your emotions, though. And in Greece it was never that big of a deal. The stoics we remember today are all Roman. Marcus Aurelius, Epicurious, Seneca.
Meditations is largely Marcus Aurelius soothing himself with logical arguments.
Im not doing it full justice but the passages read like “A whole is not less than its parts, Humans are intelligent, therefore the universe is intelligent, there for my situation is a logical plan from the universe, so I should be content.” Marcus was explaining why he needed to be the perfect elite Roman citizen, using nature to logically show why he needed to embody societally cherished traits like being logical, just, fair, but also being stern, powerful and strong. Over and over again.
Now at the time, they did consider it a hard science. They were trying to figure out secrets of the universe, the absolute correct ethics, and therapy (controlling your emotions) at the same time. Each tenet relied on eachother, you couldn’t have one with the other. Today we know the science was wrong and ethics are more complicated than they could’ve imagined, but the controlling emotions side remains. So I completely disagree that modern stoic bros are bastardizing it, it just is an flawed ideology that matches well with some of our modern western traits of a “masculine” man.
I was a fan of coheres general purpose LLM. Command A I think? Before they came out with their reasoning model.
More competition is better.
I love the design. Clean and refreshing - you start at the left and every sub element goes to the right. Like a file browser. And the commit screen is dense but super readable. I would move the summary column from the right of the README to the left of the README on the /home screen (and call /home the /README).
Loading files is very slow but I assume that's because HN is hammering the server.
I am not a believer in negative advertising. So I don't give a poop you are anti-ai. Or "better" than Github (better for who??). Just imply you are a code forge thats made for serious developers who need something engineered to be fast and reliable.
I wish you the best of luck, I can see Linear coming out with git repos after coming out with a diff reader. I have a suspicion there's space for many code forges in the market as you build out more features, especially if you lean into your products hacker-y-ness
I never thought about doing that. Thank you.
Re-read it. We needed to preserve scroll position when you went back. If you just linked back to the inbox it would kick you to the top of the screen.
I like React. And I have seriously tried the HTMX/Hotwire camp.
I wanted to make a back button use browser APIs to go back if the coming from the inbox, just link to the inbox otherwise to preserve scrolling. I had to wire the actions from the html to call the function that goes back, then in my controller determine the previous page and send the JS enabled back button or the hard link. My logic was spread out over 3 files!
With React I can have js in a component determine if the previous page was inbox, and based on that value show the back button JSX or the link. ALL IN ONE FILE. One conceptually entity for me to model vs 3 that do other things and this functionally is hammered in.
Is it slower? Definitely. But it makes me happy. Miserable in a corporate React slopbase? Blame your coworkers, it would definitely be worse without it.
If you wake up every day feeling like you got hit by a train and also have anxiety, depression, chronic heartburn, or exhaustion (not sleepiness, like a wired but tired) I would highly recommend looking into Upper Airway Respiratory Syndrome (UARS).
UARS was coined by Dr. Christian Guilleminault who was on the team that coined sleep apnea. Over his career he slowly widened his definition of sleep disordered breathing from choking (sleep apnea) to any nasal resistance that causes wake ups.
You be young and perfectly healthy but due to bad anatomy have micro arousals throughout the night. Becuase your young, your body fights through it (cuasing exhuastion) and will wake up up to readjust your airway. You won't remeber any of the wakeups, but many suffer from insomnia from it. The two big issues is large tongue in underdeveloped jaw that falls back and constricts your airway (you can test this with a mallamapati test) or a small nasal passage way.
Do you think the labs are violating their no data collection agreements for enterprises?
For reference the author's (Matt Shumer) AI business (hyperwrite ai) is a hundred small LLM wrappers that do things like:
- "transform complex topics into easy-to-understand explanations."
- "edit and transform images using simple text descriptions."
- "summarizes a research article, and answers specific questions about it."
You can see all of them here: https://www.hyperwriteai.com/aitools.
Hyperwrite does also have a markdown editor with an ai copilot sidebar that seems a little more substantial: https://www.hyperwriteai.com/ai-document-editor.
I don't know enough to disprove Matt, but I don't know why anyone should listen to him. There are far smarter people who have come up with similar conclusions.
My guess is it's based on Arch Linux. Om-arch-y.
They call it chat.
The author can't fork Rails because "the amount of work that goes into maintaining this ecosystem is enormous and expensive."
Rails IS FREE TO USE. If you want to improve test driven development, do the work yourself. Or start a company and dedicate 40% of your extremely well paid engineers time open source code others can use for free.
37signals and Shopify make the decisions because THEY DO THE WORK. I am happy to sit back and free load off of their contributions even if I disagree with DHH and Tobi's political opinions.
This happened to me over a decade ago. Medication was a godsend, and then I burned out. I remember sitting down to do work and not being able to start anything so I would pull up a dumb io game.
So I went off, and for the next 5 years I still couldn't focus. It got worse actually. I did a lot of caffeine. After COVID I started to work out and then suddenly for the first time ever I could focus. As long as I don't do caffeine, workout, and sleep I am sharp. I've done great work in the past couple years but I do feel cheated that Adderall stole time from me. I wonder where I would be with my career if I hadn't burned out.
Enough with big data! Who's working on small data? https://www.youtube.com/watch?v=eDr6_cMtfdA&pp=ygUKc21hbGwgZ...
I think it's fun that they always get in the media like this. They know how to play the game.
I've been working on an app that writes customer emails so small businesses can focus on what energizes them.
It makes answering customer emails 10x easier.
The magic are training templates which are templates that get suggested (and eventually auto-selected) and personalized by LLM for every reply.
Every reply sent trains it to auto select that training template for future similar customer emails.
The stack is Ruby on Rails and Postgres hosted on DigitalOcean. The LLM currently is Kimi K2 hosted on Groq.
Groq says it's FP16.
Grok 1 was released in 2024, Groq was founded in 2016.
No, I don't self host. I use Groq.com as my provider.
They already are. I have been using Kimi k2. It is 90% as good as Sonnet and on Groq 3x faster and 1/5th the price.
I don’t think Sam Altman said “guys, we’ve gotta vulnerable people hooked on talking to our chatbot.”
Speculation: They might have a number (average messages sent per day) and are just pulling levers to raise it. And then this happens.
This is what happens when you start optimizing for getting people to spend as much time in your product as possible. (I'm not sure if OpenAI was doing this, if anyone knows better please correct me)
Congrats on the launch! I love how you released a Linux version.
Curious about the "get smarter after every run part." Does it just feed all the past instructions into context? Or is there some RAG going on behind the scenes?
Also, I'm not an expert in marketing, but "the first desktop AI agent" doesn't resonate with me. I don't care. What does it do for me? What ever pain it solves (Something like never manually extract data from a website again) should be the header.
When a measure becomes a target it ceases to be a good measure.
The article talks about Google, but this is very prevalent in customer service too. You've had to deal with an annoying, useless chatbot.
The companies selling these chat bots advertise a resolution rate near 60%. That's a 60% reduction in work customer service has to do! But look deeper. It's not a 60% reduction in work because they're just tracking how many chats get escalated to a human. Would customers send an email to bother a human for that 60%? Probably not.
In the end customer service managers get to put "introduced AI to cut tickets by 60%" on their resume, customer experience platforms grow, and you get annoyed.
I messed up, I apologize.
I looked at the NDCG and thought that was the dataset.since voyage and cohere both used NDCG. I now realize it was separate benchmarks with the same evaluation metric.
The comment by SparkyMcUnicorn worded it better than I did.
You’re right, there’s no other way to compare embeddings than a benchmark.
Just that what the benchmark used by Voyage and Cohere tracks might not be relevant to your own needs.
Seems to under-perform voyage-3-large on the same benchmark. At the same time, I'm unsure how useful benchmarks are for embeddings.
Things are going to continue getting better, faster, and weirder.
I love this. Especially the weirder part. This tech can be useful in every crevice of society and we still have no idea what new creative use cases there are.
Who would’ve guessed phones and social media would cause mass protests because bystanders could record and distribute videos of the police?