zlatkov
Re:Coding Was Never the Bottleneck
on Monday August 31, 2026 @02:22PM (#49512993)
hahahaha
What I learned about founder-market fit
Re:Ask HN: My dev blog has 11k followers. I stopped writing it. What w...
on Monday August 10, 2026 @02:56AM (#49240221)
I actually don't think there could be meaningful offers for a blog of that scale.
Re:Ask HN: My dev blog has 11k followers. I stopped writing it. What w...
on Sunday August 09, 2026 @10:07AM (#49231565)
Yes, it was actually what got us the initial waves of high volume/quality signups to our product, so it was indeed important for the company.
Re:Ask HN: My dev blog has 11k followers. I stopped writing it. What w...
on Sunday August 09, 2026 @09:48AM (#49231402)
Good advice, and you're right that people deserve it. I never did it because I never formally decided to stop. I just assumed there'd be some form of continuity...
Ask HN: My dev blog has 11k followers. I stopped writing it. What would you do?
In 2017, I started a series on Medium called "How JavaScript works". The idea was to explain to web devs what actually happens under the hood in the browser. Some stats from the most popular articles (purely organic):* Inside the V8 engine + 5 tips on how to write optimized code: 488K views / 128K reads* An overview of the engine, the runtime and the call stack: 423K / 166K* Memory management +...
Open source "died" in March. It just doesn't know it yet
Web debugging proxy in your coding agent
Show HN: Agent Caching in Fiddler
https://www.telerik.com/blogs/stop-paying-same-answer-twice-agent-cache-fiddler-everywhere
Hi HN!We’re the team behind Fiddler, and we recently built an Agent Cache feature.It works by capturing an LLM request/response pair and, once caching is enabled, serving subsequent matching calls locally so duplicate requests never hit the provider.It’s meant to reduce three pains that show up together in agent development: token cost, feedback-loop latency, and output non-determinism.We’d...
Ask HN: What models do you use for your OpenClaw so that skills work well
Question for anyone using OpenClaw (or similar) agents daily.I've been writing new skills for my OpenClaw setup and keep running into the same thing: strong models (like Opus 4.6) do great even with complex skills, but if my OpenRouter switches to a smaller model, things just don't hold up as I have to constantly follow up, give further instructions, etc.Using bigger models constantly (via API...
OpenAI raises $110B on $730B pre-money valuation
Anyone experiencing loading problems with LinkedIn (web)?
Ask HN: How are you developing and testing agents without burning tokens?
Every agent run (and especially when running tests in CI/CD) adds token cost, nondeterminism, and slows iteration.I’m not talking about LLM evals here. I mean integration-style testing.There are some workarounds like VCR-style record/replay (e.g., Python's VCR “cassettes”) and similar ideas for capturing LLM calls. They can work for certain frameworks when your test/runtime is the one making...
Cloudflare is investigating issues with Cloudflare Dashboard and related APIs
Show HN: Agent/LLM observability for tracing, cost, evals, and debugging
Hi HN - I’m Alex, currently Head of Agent Development Tools at Progress. Before this, I was a Co-founder/CEO of a session replay startup called SessionStack, which was acquired in August this year.Since then, I’ve been pretty deep in the LLM/agent dev tools, and observability has been my main thing.I ran a small poll on LinkedIn recently about where teams are with observability for LLM-powered...
Re:Deep Dive into G-Eval: How LLMs Evaluate Themselves
on Wednesday November 05, 2025 @10:56AM (#45824302)
That's true. Even small API or model version updates can shift evaluation behavior. G-Eval helps reduce that variance, but it doesn’t eliminate it completely. I think long-term stability will probably require some combination of fixed reference models and calibration datasets.
Re:Deep Dive into G-Eval: How LLMs Evaluate Themselves
on Wednesday November 05, 2025 @10:51AM (#45824230)
I haven’t come across any research showing that a specific LLM consistently outperforms others for this. It generally works best with strong reasoning models that produce consistent outputs.