You approved Version 1 of a credit contract.
Later, Version 8 comes.
Some earlier changes may have broken your old approval.
But the part you approved still looks the same.
I've seen everything from strict 15–20 days to unlimited, and in practice the number matters less than the culture.
At places with real support and sane planning, people actually take 20–25+ days. With unlimited, it often ends up being less, because no one wants to be the first to disappear.
The worst setups are low fixed PTO with no flexibility. That just burns people out.
On my team, I actively encourage people to take time off, sometimes even without formally applying, when I can see burnout creeping in. That matters way more than whatever policy exists on paper.
What feels normal to me now is ~20 days + holidays, or unlimited with managers who actually push people to use it.
If you are enjoying vibe coding, the biggest "aha" for me is realizing that software is mostly about managing state, data flow, and failure, not syntax.
Once you get how data moves, where state lives, and what happens when things break, a lot of the "magic" disappears. Learning basic debugging and reading stack traces also compounds forever, even with AI.
One thing that helped me: think a bit like a PM. Spend time crafting good prompts (use chatgpt for generating prompts and feed it to your AI coder ;) ) for ChatGPT/Claude/Cursor, ask it to generate docs and explain what it just did. Over time, patterns emerge and things feel less mysterious.
LLMs write code well, but they don't give you mental models. That part you still have to build.
I relate to this a lot. For me, noticing the loop is already half the battle, but it doesn't mean I can exit it.
What's helped is switching from "debugging" to "externalizing". Writing the thoughts down, like logs, makes them feel less real and less recursive. Once it's on paper, it loses some power.
One small thought exercise I picked up from The Power of Now that surprisingly works for me: when I'm deep in a rabbit hole, I take a few deep breaths and ask myself, "What will my next thought be?" For a moment everything just goes quiet, and the mind kind of resets. I've used this hack multiple times.
Another thing that matters more than I expected is just building basic habits. Exercise, walking, sleeping on time, eating properly. Nothing fancy, but having a pattern makes the bad loops less frequent.
I've also learned anxiety isn't always something you can reason away. Sometimes it's just a physical state and you have to change the input before the mind follows.
I haven’t used Odoo myself, but a few friends have at small NGOs, and from our discussions your concern seems real. It works fine if you stay close to out of the box. Once you need real customization, you start depending on the partner for every small change, and costs slowly creep up.
I agree with @magnumpowers. Without someone technical in-house, it can turn into a "managed software" pretty fast :)
If most users are volunteers and non technical, I’d be a bit cautious and compare simpler tools first.
Even with good models, delivering outcomes is still messy. Bad data, unclear requirements, integrations, and blame when things break. That’s just consulting pain.
Selling access avoids delivery risk and headcount bloat. Otherwise they would become a service oriented company with AI. APIs give global distribution, cleaner margins, and optionality. They can always move up the stack later
Early on, we used Metabase for speed. It’s great for internal dashboards and quick customer access, but starts to feel clunky once you care about UX, permissions, and embedding deeply.
For customer-facing analytics, we eventually leaned toward building thin custom views on top of a metrics API / warehouse. More work upfront, but you control performance, access, and product experience.
The tradeoff is obvious, tools get you to v1 fast, custom gets you to v2+. If analytics is core to your product, I’d build. If it’s a feature, I’d buy and move on.
On small teams, AI feels like a big boost. On larger teams, it mostly just moves the work around.
We (12+ people on the team) see more LOC and faster first drafts, but also more review work. PRs look done early, but often hide shallow thinking or edge cases. Velocity goes up on paper. Review fatigue goes up too.
The best teams treat AI like a junior dev with infinite energy. Great for boilerplate and refactors, but you still need ownership. Otherwise you just ship bugs faster, which is not great honestly.
I think this shift is real, but I’m not convinced it turns into a clean new SEO anytime soon.
From building LLM systems, it feels less like ranking and more about training data, reputation, retrieval, and how easy something is to summarize. If your content is shallow or fragmented, it just doesn’t become useful to models, even if it ranks on Google.
My worry is GEO/AEO becomes the same game SEO did, people optimizing for bots instead of users. The boring strategy still wins. Write good stuff, update it, build credibility. Most tools probably won’t matter much.
So credit and risk teams check everything again.
Delta IQ shows what really needs checking. Website: https://deltaiq.tech
Feedback welcome