Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
jedisct1
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
6 ms
·
91.
▲
by
jedisct1
7mo ago
I systematically use reviewers agents in Swival: https://swival.dev/pages/reviews.html Even with the same model (--self-review), that makes a huge difference, and immediately highlights how bad the first iterations of
92.
▲
by
jedisct1
7mo ago
Codeberg is great, but I really miss Octobox. I can't imagine using GitHub without Octobox; it's just impossible to keep track of all the notifications by email. Unfortunately, Octobox doesn't support GitHub, so I've no
93.
▲
by
jedisct1
7mo ago
The US doesn't even have to ask, people already give AI providers all their data and full control over all their devices.
94.
▲
by
jedisct1
7mo ago
ugrep is my daily driver. https://ugrep.com The TUI is great, and approximate matches are insanely useful.
95.
▲
by
jedisct1
7mo ago
For open models with limited context, Swival works really well: https://swival.dev
96.
▲
by
jedisct1
7mo ago
For context extraction, Lightpanda is a really great option. Much faster than Chrome, and it comes with a built-in MCP server. However, it will not fill forms, etc. But it can be combined with agent-browser to get the best of both worlds:
97.
▲
by
jedisct1
7mo ago
Swival is really good at managing the context: https://swival.dev/pages/context-management.html
98.
▲
by
jedisct1
7mo ago
A tool like Claude Code, Codex, etc. but designed from the beginning to work well with open and local models: https://swival.dev It also comes with nice features and benchmarking abilities. For running evals, it has a companion
99.
▲
by
jedisct1
7mo ago
Qwen3.5-27B works amazingly well with https://swival.dev now that the unsloth quants have fixed the tool calling issues. I still like and mainly use Qwen3-Coder-Next, though, as it seems to be generally more reliable.
100.
▲
by
jedisct1
8mo ago
Qwen3-Coder-Next also remains amazing as a local model. If you want to use small models for coding, I'd highly recommend Swival https://swival.dev which was explicitly optimized for these.
101.
▲
Eth.zig: The fastest Ethereum library. Pure Zig. Zero dependencies
(ethzig.org)
2 points
by
jedisct1
8mo ago
|
0 comments
102.
▲
by
jedisct1
8mo ago
MCP servers were a fad, but virtually all of them are completely useless, and often counterproductive for agents that can run code and execute commands directly. When agents struggle to quickly understand how to use tools, SKILLS provide a
103.
▲
Swival – A coding agent for open models
(swival.github.io)
3 points
by
jedisct1
8mo ago
|
0 comments
104.
▲
by
jedisct1
8mo ago
Same experience here. Documentation-based skills don’t really work in practice. They tend to waste tokens instead of adding value. CLI skills are also redundant when the CLI already provides clear built-in help messages. Those help messages
105.
▲
Clojure, Reimplemented in Zig
(github.com)
5 points
by
jedisct1
8mo ago
|
0 comments
106.
▲
Zwasm: A fast, spec-compliant WebAssembly runtime written in Zig
(github.com)
2 points
by
jedisct1
8mo ago
|
0 comments
107.
▲
by
jedisct1
8mo ago
Ouch. OVH are also going to increase their prices.
108.
▲
by
jedisct1
8mo ago
They are using a wider type.
109.
▲
by
jedisct1
8mo ago
a_int and b_int are signed values.
110.
▲
Fast Sorting, Branchless by Design
(00f.net)
29 points
by
jedisct1
8mo ago
|
9 comments
111.
▲
by
jedisct1
9mo ago
Claude is good at writing code, not so good at reasoning, and I would never trust or deploy to production something solely written by Claude. GPT-5.2 is not as good for coding, but much better at thinking and finding bugs, inconsistencies a
112.
▲
by
jedisct1
9mo ago
The AirTag is a fantastic device. If only it were usable with an Android phone :(
113.
▲
by
jedisct1
9mo ago
Really cool. But how to use it instead of Copilot in VSCode ?
114.
▲
by
jedisct1
9mo ago
Related: https://00f.net/2025/07/31/mcp-as-api-wrappers/
115.
▲
by
jedisct1
9mo ago
To write Fastly VCL code, I strongly recommend XVCL https://dip-proto.github.io/xvcl/ It makes VCL so much easier and readable.
116.
▲
by
jedisct1
10mo ago
Because the authors found out about it by chance on Hacker News. That said, these issues are not a big deal. The first one concerns someone manually reading a signature with cat (which is completely untrusted at that stage, since nothing ha
117.
▲
Wallbleed: A Memory Disclosure Vulnerability in the Great Firewall of China
(gfw.report)
6 points
by
jedisct1
10mo ago
|
1 comments
118.
▲
by
jedisct1
10mo ago
If a website can crash your browser, the problem is your browser...
119.
▲
by
jedisct1
10mo ago
Doesn't seem to work with Zig?
120.
▲
by
jedisct1
10mo ago
LLMs don't need your secret tokens (but MCP servers hand them over anyway): https://00f.net/2025/06/16/leaky-mcp-servers/ Encrypting sensitive data can be more useful than blocking entire requests,
More ›