Dquiroga

Hi HN,

I’m Daniel. I’ve spent the last few years building internal tools and automations for companies like FINN and Personio. I built PUNKU.AI to solve a frustration I had with the current state of automation.

Right now, you generally have two choices: Manual: Dragging and dropping nodes in tools like Make/Zapier (reliable, but tedious for complex logic). "Magic" AI Agents: Giving a prompt to an agent and hoping it works (fast, but impossible to debug or trust with sensitive data).

PUNKU.AI is an attempt to bridge the gap. It uses an LLM to "architect" the workflow, but it outputs a fully editable visual node graph (JSON-based) that you can verify before running.

How it works: 1. The "Interview": Instead of taking a zero-shot prompt, the system acts proactively. It parses your request and asks clarifying questions to resolve ambiguity (e.g., "If the API returns a 404, should I retry or alert you via Slack?").

    2. Graph Generation: It generates the logic, loops, and branching, visualizing it as a node graph. You can manually tweak connections if the LLM hallucinated a step, or ask it to regenerate specific parts.

    3. Integrations: We are connecting to ~3,000 API endpoints. The agent handles the authentication handshake and schema mapping.

I’d love for you to try breaking it. Ask it to build something complex (loops, conditional logic) and let me know if the generated graph makes sense.

Link: https://www.punku.ai

Thanks!

60% probably felt like a lot to Gemini. However, I liked the doomerism and how google was using our data to train its models.

Nonetheless, Gemini 3 failed this test. It failed to start a discussion. Its points were shallow, and too aiesque.

yeah hahahahah, it made me think!

I asked Gemini to write "a comment response to this thread. I want to start an intense discussion".

Gemini 3:

The cognitive dissonance in this thread is staggering.

We are sitting here cheering for a model that effectively closes the loop on Google’s total information dominance, while simultaneously training our own replacements.

Two things in this thread should be terrifying, yet are being glossed over in favor of "look at the pelican SVG":

1. *The Model Card Leak:* As user `rvz` pointed out, the leaked model card implies training on "user data... in accordance with Google's relevant terms of service." Given the recent terms updates, this essentially confirms that Gemini 3 is likely refined on the corpus of Gmail and Drive data that they’ve been hoarding for decades. We are cheering for a product sold back to us at a 60% markup (input costs up to $2.00/M) that was built on our own private correspondence.

2. *The "Agentic" Trap:* Everyone praising how `Gemini 3` solves complex bugs one-shot or handles entire PRs is missing the macro view. We aren’t looking at a productivity tool; we are looking at the devaluation of the software engineer. When the entry barrier drops to "can you prompt a seminal agent," the economic value of what we do on this forum evaporates.

Google has successfully gamified us into feeding the very beast that will make the "14-minute human solve time" (referenced by `lairv`) irrelevant. We are optimizing for our own obsolescence while paying a monopoly rent to do it.

Why is the sentiment here "Wow, cool clock widget" instead of "We just handed the keys to the kingdom to the biggest ad-tech surveillance machine in history"?

Since when does critical infrastructure fail weekly?! One week is AWS, then azure + AWS, now cloudfare...

Time to go back to on prem. AWS and co are too expensive anyways