Syntaf

8 ms·

1.
by Syntaf 3d ago
> It's a place for assholes Sounds like you're contributing to your own narrative here. The only self-centered douche canoe of a comment I see in this thread is from you. HN has its share of good & bad, but I've genera
2.
by Syntaf 10d ago
Yeah this article is quite shallow after having read it; delegating to a remote sub-agent sounds novel but how does it _actually_ perform? I'm personally skeptical of optimizing for minimal token consumption, the closer to a vanilla se
3.
by Syntaf 13d ago
I've almost entirely stopped using Opus 5, at least with Sonnet I know what i'm getting -- Fable for the complex stuff and Sonnet for the precision changes. The verbosity of Opus 5 isn't even my issue, it's consistency.
4.
by Syntaf 20d ago
I think this article posted to HN last week did a better job explaining things IMO https://earendil.com/posts/what-is-a-harness/
5.
by Syntaf 21d ago
One of my favorite articles posted here for this year, it's a great read and really gives you an idea of just how dysfunctional azure is.
6.
by Syntaf 23d ago
I don't have a strong argument for or against using MCP, it honestly comes down to familiarity. In my own opinion, a CLI tool is going to be much more familiar ground for engineers -- I wouldn't expect the 200+ engineers at my com
7.
by Syntaf 24d ago
Great article and super relevant advice. As building becomes easier and easier in our career those with the agency to find problems worth solving become x10 more productive. Like other people in this thread are also saying, the more you mas
8.
by Syntaf 24d ago
Sorry I can't share what we're doing here directly! I will however say that this page alone does a pretty good job of illustrating what an agent harness might look like: https://docs.agno.com/tools/overview *
9.
by Syntaf 24d ago
It’s actually encoded in an abstraction that we call “gates” which run before any tool call an agent makes, this allows us to prevent the tool call from happening and return a cited code + explanation on why their tool call was not executed
10.
by Syntaf 24d ago
Yeah the CLI can provide schema for commands via the usual ‘—help’ syntax, so agents are able to discover + explore commands on their own. As for an example: if our agent wants to book a journal entry to cash accounts for a client, it MUST
11.
by Syntaf 24d ago
I’ve been working on a harness for accounting agents at my job recently and it’s been a pretty interesting experience. We originally started with building a CLI tool so our LLMs could more easily interact with our platform. I cannot recomme
12.
by Syntaf 27d ago
With a whole separate LLM? FWIW you can also consider using something deterministic like vale.sh [1] , it's pretty easy to write prose violation plugins [2] [1] https://vale.sh/ [2] https://github.com/S
13.
Hop: CLI tool for jumping around worktrees(github.com)
2 points by Syntaf 1mo ago|1 comments
14.
by Syntaf 1mo ago
TL;DR `hop claude/57680` will cd you into the worktree for the branch. I've wrote this tool so I can do things like `hop claude/57680 && vim index.ts` and quickly edit a file on a branch without having to ask claude t
15.
by Syntaf 1mo ago
Totally agree with your framing here, but that is also what I fundamentally consider "good" code -- it's code that solves the problem that your customers/business needs without making it _harder_ to solve the next proble
16.
by Syntaf 1mo ago
Fair point! I edited my comment, A&D -> Architecture & Design
17.
by Syntaf 1mo ago
> bad engineers were always a liability This part of the article hits home for me. With AI, "bad" engineers can now amplify their "bad" engineering x10 across the organization. The most egregious of these cases for me
18.
by Syntaf 1mo ago
My company is migrating repositories over to a self-hosted forgejo instance, currently very jealous of the engineers working on those migrated repos.
19.
by Syntaf 1mo ago
There was a great article awhile back that shed some light on just how dysfunctional azure is as a platform: https://news.ycombinator.com/item?id=47616242 If I had to guess it's because Github is sitting on top on infr
20.
Vale-LLM-slop – prose linting for LLMs(github.com)
2 points by Syntaf 1mo ago|1 comments
21.
by Syntaf 1mo ago
Hey HN, recent hype around the ASD-STE100 standard inspired me to think about a more deterministic way to lint my agent's prose. This vale style is inspired by the english standard and is an attempt at providing a standard for teams to
22.
by Syntaf 2mo ago
So let me get this straight, in this article we have: - One guy selling shovels (AI “tools”) to clueless people - Another guy who started and then folded his business in two months - A chatGPT wrapper (it’s literally called chatPRD) This is
23.
by Syntaf 2mo ago
Does this actually work for you? I've tried adding rules in my user profile as well as adding instruction in my CLAUDE.md but my models always eventually go off the rails and start spewing technical jargon in doc strings & descript
24.
by Syntaf 2mo ago
It’s the pineapple on pizza of ops, people just like to fit in sometimes. I’ve been running my own personal k8s cluster on digital ocean for the last 5+ years now and it’s dead simple. Takes me 30 minutes to create a new namespace a deploy
25.
by Syntaf 2mo ago
There’s got to be more to this story, what exactly were you up to with these models?
26.
by Syntaf 2mo ago
Ok long time Claude Code user here; lately I've started to realize there's other great models out there I should be trying, but I'm hesitant to leave Claude Code behind for something new. What's the consensus today on co
27.
by Syntaf 2mo ago
Both very fair observations. > Whether this mode of working is going to be long-term viable is going to depend on how important it is for you to be aware of what has happened for the system in question This is the million dollar we'
28.
by Syntaf 2mo ago
Does Joel still disagree today? Worth noting that this article is 25 years old. The world was very very different back then, especially when it comes to software engineering. Context switching is a problem when the cost of switching context
29.
by Syntaf 3mo ago
I used the chrome MCP to profile a slow react page at my company, set a /goal and had it iterate until it achieved under 100ms responsiveness to actions. Claude was able to identify the slowness and use various react tricks to fix the
30.
by Syntaf 3mo ago
6.x feels much more efficient with respect to token usage to be fair. I picked up superpowers back when it first started gaining traction; the first iteration felt like an “oh shit” moment for me, then the sheen quickly wore off. Higher spe

More ›