hedgehog

9 ms·

181.
by hedgehog 5mo ago
Most of mine have code in them. That's most of the value.
182.
by hedgehog 5mo ago
I've been using various UNIXes and clones since the 90s so I do generally know what's going on but I also have no desire to fill my brain with the syntax to the new new new commands to configure an Ethernet interface on Linux etc,
183.
by hedgehog 5mo ago
Claude Code + Ansible + whatever stuff needs to be managed gives some visibility and control and in my experience is reliable enough to be useful.
184.
by hedgehog 5mo ago
Multi-token prediction is available now, I'm still getting it set up but it sounds like it'll be 1.5x or 2x on the bigger models.
185.
by hedgehog 5mo ago
If you want another take from that period the 1997 book "Apple" by Jim Carlton (WSJ reporter) is pretty good. The inside jacket starts "Whatever happened to Apple Computer?" and the forward by Guy Kawasaki frames the boo
186.
by hedgehog 5mo ago
I think the ETA thing did work at one point.
187.
by hedgehog 5mo ago
My suspicion is a lot of the difference in performance in newer models comes from more and better code reasoning and debugging tasks in the RL phase, along with actual security bug finding workflows. When sessions get long and instruction-f
188.
by hedgehog 5mo ago
I hadn't taken a look and indeed submitting bugs into the Firefox bug program looks much more accessible than upstreaming a patch into an open source project. It's true just asking the robots "find bugs" isn't enoug
189.
by hedgehog 5mo ago
Me, driving, using Apple Maps for navigation: "Hey Siri, what's my ETA?" Siri: ... "Here's what I found on the web for "what is my ETA'" From outside I don't know the cause but contrary to their
190.
by hedgehog 5mo ago
I did some experiments and Opus seemed pretty able to wire up a harness to find bugs and write PoC + patch for each. It's still a lot of work to get fixes upstreamed from outside so I think even if outsiders have better tools (Mythos e
191.
by hedgehog 6mo ago
They used repeaters on street lights as part of the infrastructure, and even after the company went belly up people were able to use the repeaters for private networks. Pretty slick for the mid 90s.
192.
by hedgehog 6mo ago
In my opinion maintainers can easily run a "hey robot, scan my code for risky patterns" to get a rough list, or they can solicit unreviewed contributions, but otherwise better not to add noise.
193.
by hedgehog 6mo ago
It often takes strong understanding of the upstream codebase and roadmap to write a good patch. It's easy enough to write a rough PoC and draft patch but getting all the way through the cycle takes up a bunch of time both from you and
194.
by hedgehog 6mo ago
As far as I can tell Minimax M2.7 is better than anything available a year ago, but it runs on an ordinary PC. Will that continue? Not sure, but the trend has continued for the last two years and I don't know of any fundamental limits
195.
by hedgehog 6mo ago
I have six patches that I will at some point upstream, the main bug/surprise is the .gitignore behavior is not what's documented, but even without it seems to work quite well.
196.
by hedgehog 6mo ago
I'm not sure one way or another but I've been using a related tool called Tilth by another poster here. It doesn't do anchor-based editing, but it does do syntax-aware search and will e.g. report the line range for function d
197.
by hedgehog 6mo ago
It depends where you prune and how the specific prefix cache you're targeting works. Pruning or condensing recent items that are unnecessary probably pays for itself.
198.
by hedgehog 6mo ago
How do you benchmark that? The problem with extracting memories in the background is it's hard to make that work with the prefix cache. You can go a long way with a simple 2-stage LOG.md (detailed log of tasks & lessons) + MEMORY.m
199.
by hedgehog 6mo ago
I used Opus via Copilot until December and then largely switched over to Claude Code. I'm not sure what the difference is but I haven't seen any of these issues in daily use.
200.
by hedgehog 6mo ago
Have the tool maintain a doc, and use either the built-in memory or (I prefer it this way) your own. I've been pretty critical of some other aspects of how Claude Code works but on this one I think they're doing roughly the right
201.
by hedgehog 6mo ago
It would be funny to do an optimization pass to find a compact description of how to coax an accurate pelican bicycle out of a few of the current models, then just blast that snippet everywhere.
202.
by hedgehog 6mo ago
I have historically not used them but I want to start so that some of the spin up / tear down work of doing any particular task can happen in isolation. For example, drafting a change before I start editing, checking out and setting up
203.
by hedgehog 6mo ago
If I did a proper benchmark I think the numbers would be what you got. Minimax M2.7 is also surprisingly not that slow, and in some ways faster as it seems to get things right with less thinking output. (around 140 t/s prefill and 23 t
204.
by hedgehog 6mo ago
This one is around 250 t/s prefill and 12.4 generation in my testing.
205.
by hedgehog 6mo ago
Please describe the process as a program written in standard Scheme.
206.
by hedgehog 6mo ago
The article doesn't really explain the overall budget, for scale it looks like in the 2025-2026 budget year CA planned to spend about $228B compared to $216B revenue ($227B in the previous year). https://ebudget.ca.gov/
207.
by hedgehog 6mo ago
Similar path, I was very skeptical but Claude touches over 90% of my code now. I review almost all of it though. I did not have high expectations for Claude Design but yesterday I tried it for a workflow tool I'm building and in one sh
208.
by hedgehog 6mo ago
That's a very insightful observation, highlighting the genuine tension between consistent messaging and quick, pithy responses on Hacker News.
209.
by hedgehog 6mo ago
I have some practices that seem to be very effective but I wouldn't call them "vibe coding" and they're not really exciting enough to write a blog post about. Mostly: code is cheap, review labor is not, write down your p
210.
by hedgehog 6mo ago
Did you end up finding a reference you liked better? I'm default suspicious of anything called "vibe coding" but there are probably some good lessons in that territory.

More ›