osigurdson

6 ms·

61.
by osigurdson 4mo ago
It hurts, but I'll pay a premium for worse hardware that will run Linux. But for others in my family Macbooks are just far better value by a wide margin than Windows laptops. Vertical integration really wins these days.
62.
by osigurdson 4mo ago
Automated outbound is a real problem now. I think LinkedIn actually has a pretty good model in which it costs money to send a message if the other party does not reply. Lucky for them of course! Nice to be able to collect 100% of the value
63.
by osigurdson 4mo ago
Does containerization help much here? If it's a code tool then presumably it needs access to your code files (read / write). Maybe there are use cases for it of course.
64.
by osigurdson 4mo ago
Right. You mean all behaviors are tested, essentially. So if you / team are going to implement a new feature, what does that look like? Do you write Gherkin or similar, unit tests or both? Can you provide an example of what that might
65.
by osigurdson 4mo ago
Back in the crypto days I was somewhat intrigued by Arweave's permanent storage idea. It seems like it is still around (I've never used it personally though).
66.
by osigurdson 4mo ago
>> if your tests don't catch any known issues, the problem is the tests aren't comprehensive enough. Maybe I misunderstand but this seems like a fairly low bar in the test suite only covers existing bugs. I'd argue tha
67.
by osigurdson 4mo ago
Definitely agree that performance optimization is a good use case for LLMs. Here you have both a measurable goal / objective function and guardrails against functional regressions. It kind of closes the loop in that regard. One thing h
68.
by osigurdson 4mo ago
I can see how you could avoid regressions this way, but what do you add to your harness to prove that a new feature is working?
69.
by osigurdson 4mo ago
Exactly. Popular opinion is behind reality by several months. Claude used to be significantly better, now it is basically the same.
70.
by osigurdson 4mo ago
>> MCP consumes ~65x more tokens than the CLI approach. For this example, there seems to be no explanation for the LLM to know when to use this curl command, etc. Is the idea that the linear API is known in the LLM weights already and
71.
by osigurdson 4mo ago
Maybe three days is enough for some code bases, but if you have millions of lines of code, agents aren't going to help you that much.
72.
by osigurdson 4mo ago
I remember it was like that but it was kind of silly. You had "Software Engineers" working alongside "Software Developers" doing exactly the same job. The purpose of the P.Eng. was to protect the public, not people'
73.
by osigurdson 4mo ago
While I think a dead economy is easy to imagine it is also pretty absurd. After all, real supply will be abundant and real demand will be strong except since so one has any (artificial) money, everything presumably stalls and we somehow end
74.
by osigurdson 4mo ago
I don't think you can call yourself an accredited civil, mechanical, etc., engineer, even in the US. However, you can call yourself a software engineer nearly anywhere because it isn't the same thing - no regulatory body is going
75.
by osigurdson 4mo ago
Here are a few job descriptions for reference: https://news.ycombinator.com/item?id=47975744 https://news.ycombinator.com/item?id=48099785 https://news.ycombinator.com/item?id=47978246
76.
by osigurdson 4mo ago
If those are CDN urls then likely they resolve to IPs within the EU so that use case would probably be fine. >> throwing up warnings pretty often That is mostly the situation (globally) with cookie banners. There would be a transition
77.
by osigurdson 4mo ago
I don't think this explains the phenomenon as is more temporal in nature - not prompt to prompt. I'm sure the AI labs gracefully degrade to simpler models when resources are low - why wouldn't they?
78.
by osigurdson 4mo ago
ClickHouse is fine for looking at recent data (with simple / efficient TTL). I'd still (probably) use Postgres for smaller scale things however.
79.
by osigurdson 4mo ago
From experience, I'd suggest using ClickHouse beyond a few billion rows of timeseries data in Postgres.
80.
by osigurdson 4mo ago
I find the quality ebbs and flows even on the same model. My guess it is something to do with GPU availability but only guessing.
81.
by osigurdson 4mo ago
Presumably, EU citizens want EU regulations for privacy. I'm suggesting that an EU browser merely provides a warning when a non-EU regulated site is being visited. So, the user is in control and it avoids the scenario where the interne
82.
by osigurdson 4mo ago
I think the verification would start at the origin site - wherever the page located. All traffic originating from the client side bundle or on the server would have to stay in the EU. Cloudflare is going to provide you with a local IP for w
83.
by osigurdson 4mo ago
>> UK is not in the EU, so you'll need another browser. I'm thinking it would just be a warning. The EU citizen can then decide if they want to take on the associated risks of using a non-EU compliant site.
84.
by osigurdson 4mo ago
Does this mean that any site operated by a US company is automatically in violation of EU regulations? If so, the EU browser could logically filter these out as well. Basically the EU browser should only work for sites that are compliant wi
85.
by osigurdson 4mo ago
These websites would logically have to move to EU based servers in order to be subjected to the kind of regulation that EU citizens want. I think it could work.
86.
Ask HN: Why not have an EU browser?
7 points by osigurdson 4mo ago|20 comments
87.
by osigurdson 5mo ago
I think it would be better if there was an EU browser that provided a warning if accessing sites outside of the EU. If an origin site is in the EU then it could be subject to EU regulations.
88.
by osigurdson 5mo ago
I don't see search and AI as fundamentally distinct things. Usually I just want an answer.
89.
by osigurdson 5mo ago
Realistically, OpenAI found product market fit with the OpenAI API playground in 2021. People were using that as ChatGPT at the time.
90.
by osigurdson 5mo ago
What concrete difference is being made here? If a site is hosted in the US but accessible by EU citizens AND using a PostHog US server isn't the data still being used for training? Legitimate question, I am not trying to prove a point.

More ›