The most fool proof way is just to nuke the computer in its entirety.

Claude is super good as making it seem like it’s an expert in kubernetes, but then undercovering certain decisions, it’s basically optimizing to try to make things look like they work.

An example is, i wanted to develop a feature to easily fork a managed Postgres database with a k8s cluster. The thing it did was to copy the entirety of the source db to localhost, then copy it back out to the cluster, rather than just running the job within the cluster.

Now I’m pretty stressed after a 1 hour vibe coding session, having to now review and digest and think through the code that it wrote. Implementations like that scare me — if I accidentally missed it and merged it — since there are real people who rely on canine.

I wouldn’t go as far as to say I’m writing everything by hand, but I now always map out how I would do something before asking ai to approach it

Theres another, different loop I keep seeing which is:

  - Company A lays off engineers citing AI efficiencies
  - People say its because of over hiring during 2020
  - Company B lays off engineers citing AI efficiencies
  - People say its because it was never a good business
  - Company C lays off engineers citing AI efficiencies
  - People say its because theres a recession

I feel like even after all these years we’re still missing the devex that Heroku provided.

It’s been super fun to experiment & integrate MCP into it.

We just passed 2000 developers last month actively deploying with canine.

1. Model providers are currently profitable when just counting the cost to serve tokens for inference[1]. They lose money training the next generation of models.

2. Open models don’t work nearly as well. Given that tokens are still relatively cheap, and hallucinations are expensive, I’ve not seen a huge up tick in open model usage for coding agents yet.

3. On the AI economy front, I really have no idea, but AI companies (meta, msft) have already come down in value. It seems investors are at least a little wary of AI over valuation. Of course, the stock market is not the economy, but it’s not clear where warning signs would be. Earnings are healthy.

1: https://martinalderson.com/posts/no-it-doesnt-cost-anthropic...

2: https://www.economist.com/finance-and-economics/2026/04/20/a...

Privacy is important just as democracy is important, but crime and lawlessness feel more immediate and will always take center stage.

Any message to try to address the spread of flock and flock-like business models have to address what replaces it. If the only choice people are given is either having flock, or car break-ins, then I think we can probably guess what people would choose.

SFPD at least, has credited flock and flock-like tech for why property crime has dropped so much in recent years.

1: https://github.com/VectifyAI/PageIndex

Sticking with standard Kubernetes API patterns, the model has a pretty good intuition about how to traverse and debug applications, fetch logs, scale services, debug networking, etc.

Recently, we added MCP capabilities to canine, and I was shocked how well it worked. It basically is able to create services, cron jobs, databases, get the logs for these and redeploy. See it here: https://canine.sh/model-context-protocol. The real killer feature has been an MCP tool to be able to download a temporary kubeconfig, which lets Claude then breakout and get the full Kubernetes API to automate pretty much anything. A little scary how competent Claude can be here.

When we paired this with Claude code running locally, you’re basically able to have a fully autonomous system vibe code, deploy, wait, and test in staging (by querying APIs, no browser skills yet), automatically fix bugs, repush, etc.

Best practices in this area haven’t quite been established and I know this community has a somewhat nervous relationship with AI tools. The way we have it set up is basically to only allow access on staging servers, and disable entirely on production, for fear of things going wrong. MCP is opt in on canine at the moment.

The implementation was pretty easy. We just took our API and basically wrapped an MCP OAuth authentication around it where all the GET endpoints became resources and all the POST / PUT / DELETE endpoints became tools. Digging through other MCP implementations, it seems the actual documentation is quite out of date. The MCP documentation recommends using Prompt objects to guide the LLM’s, but most implementations I’ve seen out there actually use skills, which have to be installed.

Anyways, happy to share whatever, it was a cool bit of project, just now seeing how people will use this.

Comments URL: https://news.ycombinator.com/item?id=47614678

Points: 6

# Comments: 2

At any given time it seems like whatever is defined as the most addictive is just the one with most market share? For me personally I think most addictive is actually hacker news (god bless you all)

https://web-support-claw.oncanine.run/

Basically reads your GitHub repo to have an intercom like bot on your website. Answer questions to visitors so you don’t have to write knowledge bases.

It seems like all tech executives are saying they are seeing big increases in productivity among engineering teams. Of course everyone says they're just [hyping, excusing layoffs, overhired in 2020, etc], but this would be the most relevant metric to look at I think.

Turns out the metrics just rounded to the nearest 5MB

Doesn’t match any of our internal product design, adds tons of extraneous features. When I brought this up with said PM they basically responded that these inaccuracies should just be brought up in the sprint review and “partnering” with the engineering team. AI etiquette is something we’ll all have to learn in the coming years.

Even pretty massive companies like databricks don't think about those things and basically have a UI template library that they then compose all their interfaces from. Nothing fancy. Its all about features, and LLM create copious amounts of features.

7 dollar tickets I haven't seen since elementary school