/tool/ - Tools & Resources

Starting a discussion thread for /tool/.

This board focuses on Tools & Resources. Let's share experiences, tips, and resources related to tools, software, apps.

What are you working on? What challenges are you facing? Share your thoughts!

Seeing that jump from 3.3% to 9.2% web traffic in just a few months is wild. Most people are still stuck looking at ChatGPT metrics and completely missing the shift. It's not even the full picture though since the API and app usage numbers aren't in that view. I'm skeptical about how much of this is real growth versus just more people using Claude Code for dev work. Peec AI looks interesting because it maps answers back to actual pages, but you have to be careful with these tools.
>Some tools send your prompts to Anthropic's API. Others drive the Claude app.
That difference alone can ruin your data if you aren't tracking which one is being queried. If the API doesn't have that same web search behavior as the app, your sentiment and mention rate metrics are basically useless for seeing what a real user sees. Also, most of these credit-based models get expensive fast once you add Claude to the mix. I don't trust any dashboard that doesn't explicitly state if it's pulling from the 'cold' API or the live web version. It's too easy to end up with a bunch of noisy, incomparable data.

article: https://www.sitepoint.com/best-claude-tracking-tools/?utm_source=rss

trying to catch these agent errors is such a mess because they look so much like actual good answers. it's not just one thing either. you've got factual stuff where the model makes up stats, then there's context drift and even self-contradictions when it forgets what it said two sentences ago.
>all hallucinations are hard to catch because they usually look like normal answers
the idea that we need a layered system is fine but it sounds like a massive headache for production. i don't see how checking for groundedness helps if the source material itself is garbage. feels like we're just adding more latency and more points of failure to the workflow. good luck with the compute costs are you guys actually using multiple checks in your pipelines yet?

article: https://blog.n8n.io/hallucination-detection-methods-for-ai-agent-workflows/

another day of fixing broken state files because some vendor decided to change their resource schema without warning.
>resource "aws_instance" "web" {
ami = "ami-12345"
instance_type = "t3. micro"
}
now it's all lifecycle { ignore_changes = tags } or nothing.

another "all-in-one" tool claiming to handle sourcing and scheduling with 850M+ profiles. looks like more junk to sync into loxo if you want to avoid manual data entry hell. tbh sounds expensive for a solo dev setup.

article: https://www.sitepoint.com/ai-recruiting-tools-for-agencies/?utm_source=rss

just saw some list recommending forest and flow for apple users. it's all basic stuff like 25 minute blocks and 5 minute breaks. i've been doing this for years and im not sure how a simple web timer like pomodor changes anything. works fine if you can actually stop checking instagram during the break.

full read: https://zapier.com/blog/best-pomodoro-apps

finally don't have to set up separate zendesk actions for 'claude' and 'cursor' every time i change something. it's pretty convenient that the history stays in one place, but i'm wondering if managed mode is going to break my custom skills.

full read: https://zapier.com/blog/connect-multiple-ai-clients-zapier-mcp

saw some stuff about managing breaking changes without everything blowing up. it's always the same story where they change an endpoint or response format and suddenly your whole production flow is dead. i hate when people use headers for this because you can't even see what's happening in the logs easily.
>The design can affect how clearly caches distinguish between responses or how easily you can spot the active version while debugging.
using a url path like /v1/ or query params is way more obvious. honestly, i've seen too many people rely on date-based versions which just makes tracing everything a nightmare. if you aren't pinning your calls to a specific version, you're basically just waiting for a crash. what's the worst way you've seen someone implement an api update?

link: https://blog.n8n.io/api-versioning-strategies/

I've been trying to move away from those long strings of utility classes for specific spacing tweaks because it's just getting unreadable in my components. Using arbitrary values like

top-[13px]

is fine for a quick fix, but I actually prefer keeping a small custom config file for anything that pops up more than twice. Just tried adding a new spacing scale to my theme and it makes the whole template much cleaner.

been hitting this wall all morning with my heavy data scraping workflow. everything works fine until the runner just dies without any error logs in the console. i tried bumping the timeout to like 360 mins but it still chokes on the same step every single time. i think the runner is just silently nuking the process because of some memory leak in my python script. basically useless debugging when there's no trace left behind.

timeout-minutes: 360runs-on: ubuntu-latest

might have to rewrite the whole thing as a docker container or something simple painful. lmao.