haxton
Joined 9 years ago
Karma
138
Submissions
40
Member since
Mar 29
unlicensed plumber who occasionally writes code
View on Hacker NewsRecent submissions
Comment · 11 months ago
Why is this AI generated slop so highly upvoted?
Comment · 1 year ago
California has banned non competes since 1872. You might be thinking about non solicits which was 2024 also reaffirming the ban on non competes
Comment · 1 year ago
Unrelated. Has nothing to do with the gcp outage that was related to.
Comment · 1 year ago
Always treat startup equity as 0 until you've sold it.
Comment · 1 year ago
> Custom OAuth implementation with user registration/login
Please don't. At 2 minute glance you are allowing empty state (csrf) and bearer tokens in query string[0], not checking if token is expired or not[1], storing secrets in plain text / not salting[2], missing PKCE Validation, debug mode always on, redirect URL only checking if includes (127.0.0.1.evil.com works)[3] so much...
Please, please, please don't recommend this for any production usage.
[0] https://github.com/f/mcp-cloudflare-boilerplate/blob/master/... [1] https://github.com/f/mcp-cloudflare-boilerplate/blob/master/... [2] https://github.com/f/mcp-cloudflare-boilerplate/blob/master/... [3] https://github.com/f/mcp-cloudflare-boilerplate/blob/master/...
Comment · 1 year ago
> The former Google employee
It's literally like the 3rd sentence...
Comment · 1 year ago
I've been about it as "throwaway software." Why bother searching for someone else's mediocre LLM generated software when I can just as easily (and hopefully as cheaply) generate the same thing, but it just works for me
Comment · 1 year ago
Kindle has been doing this for years and has really made me a loyal customer to them. Always surprised the penny pinchers at Amazon haven't killed it yet.
Comment · 2 years ago
The demos I see for these types of tools are always some toy project and doesn't reflect day to day work I do at all. Do you have any example PRs on larger more complex projects that have been written with codebuff and how much of that was human interactive?
The real problem I want someone to solve is helping me with the real niche/challenging portion of a PR, ex: new tiptap extension that can do notebook code eval, migrate legacy auth service off auth0, record and replay API GET requests and replay a % of them as unit tests, etc.
So many of these tools get stuck trying to help me "start" rather than help me "finish" or unblock the current problem I'm at.
Comment · 2 years ago
Less of a HN issue and more google allowing UGC on their research subdomain
Comment · 3 years ago
My favorite example is the asana loader[0] for llama-index. It's literally just the most basic wrapper around the Asana SDK to concatenate some strings.
[0] - https://github.com/emptycrown/llama-hub/blob/main/llama_hub/...
Comment · 3 years ago
gpt3.5 turbo is (mostly likely) Curie which is (most likely) 6.7b params. So, yeah, makes perfect sense that it can't compete with a 70b model on cost.
Comment · 3 years ago
Elbow method is a good place to start for finding the number of clusters.
Comment · 3 years ago
Curious to know what value you've seen out of these clusters. In my experience k means clustering was very lackluster. Having to define the number of clusters was a big pain point too.
You almost certainly want a graph like structure (overlapping communities rather than clusters).
But unsupervised clustering was almost entirely ineffective for every use case I had :/
Comment · 3 years ago
GPT4[0] is actually very good with base64 to the point where it makes perfect sense.
I'd be interested in how well you could finetune 3.5 to use different compression.
[0] - https://platform.openai.com/playground/p/hfLUBCTE8RrRYPIRxEe...