AskBox

Lalitha A R

AI · Research · Tooling · Design / UX · Data & Analytics · Software Engineering · Product Management · Marketing

Anonymous asked:

We’re considering moving away from Webflow for something more AI-friendly. Any recommendations? Our biggest pain point today is the MCP experience: Slow, Browser-dependent, Token-hungry. What I’m looking for is pretty simple: Have an excellent building and editing experience with AI. For me or any person of the team. But I don’t want to spend any time thinking about Core Web Vitals, technical SEO, or infrastructure. Curious to hear what you’re using and what you’d recommend 🙏

A mix of custom and switching to Quarto for documentation/static sites? But that wouldn't do without thinking of technical stuff, yes.

Maybe try to map out core stuff you keep altering, any patterns? If you notice any patterns, it can be a one-time heavy build, but with proper documentation, that build will help the AI build it further without adding overhead on the technical part.

Given it's not a really niche tool and something AI is familiar with and well documented, Claude will help you manage it a lot.

I know tech stuff may sound overwhelming (and it is, even for devs — things keep changing), but if we trace back to common patterns and nail down what helps, the day-to-day overhead becomes really manageable and tweakable with Claude.

Web dev isn't what I do as a main thing, it's research, and I'm taking a break from research right now. I have enough tech to set this up and enough of the non-tech side to know what's sustainable without tech overhead. It's really distracting to manage too much technical stuff when your core remit isn't tech at the moment.

For context: we had a site that was really cool and didn't take much time from core research, but the moment we had a new announcement, blog post, or feature, the whole thing took hours and became a huge chore. We traced back the last month of updates and what we expected to change in the next three, and also needed it to look on-brand, so we couldn't just slap a template on it.

If there's one thing I'd suggest, it's pattern mapping. A site usually has few moving parts, and few that only matter when there's a big decision. Tracing that helps you see what's actually consuming time, so you automate that instead of bending yourself around an existing system.

A few things worth checking regardless, for the first build:

1. When does it break (not every break is a KPI — if it breaks and it matters, fix it).

2. Does it follow DRY, is there a coherent structure, or is it spaghetti?

3. A documentation version for a future AI agent, and a separate one for a future non-technical human — they read very differently.

4. A "repo map" markdown doc, so a future agent doesn't burn tokens figuring out where things are.

5. An authority matrix — what an AI can edit outright, what needs a flag to a human, what needs a technical consult.

6. Tests, so future updates don't quietly break what already worked.

Anonymous asked:

When discussing once you told me to go for primary sources when collecting data, aka I have to look if a statistic/claim is empirical to a paper or if they are citing some other paper as the source of that claim. But I come across really good awareness organization's websites, they will have verified before using a statistic right? Why can we not cite them?

We have a lot of respect for these organisations and the work they do. The reason they are not citable here is not about credibility — it is about purpose and audience. Awareness organisations write for the public. Their goal is to persuadeView more

Anonymous asked:

I am working on research where I am reading a lot of different country's law on a particular topic. But I am struck - how will I know whether what I find is the most recent updated law? I tried using the most recent once as available on the site but turns out there was an amendment just few weeks back that wasn't updated in the page but was available in the news and a court case, I almost missed it.

First, go to the official regulatory body's website directly — not a third party that hosts the document. If we are talking of Indian Food safety, it is fssai.gov.in, for the EU it is eur-lex.europa.eu, for FSANZ it is foodstandards.gov.auView more

Anonymous asked:

My background is quite different from many researchers here. I have spent over 30 years in industrial biotechnology, focusing on fermentation, bioprocess development, manufacturing, and process scale-up. I'm now exploring how AI can support scientific decision-making in industrial R&D. One thing I've found particularly valuable is using Claude not only for literature review, but also for: Comparing experimental results across multiple studies Identifying conflicting conclusions in the literature Designing experiments and DOE plans Reviewing SOPs and suggesting improvements Brainstorming hypotheses before moving to laboratory validation I'm still learning the best ways to integrate Claude Science into an industrial research workflow, so I appreciate posts like yours. I'm curious—have you found any prompting techniques or workflows that significantly improve the quality of scientific reasoning, beyond standard literature summarization?

Thank you :> I haven’t tested Claude Science extensively. Incase of general Claude, here are some things that work really well, - If you are indeed using it to find literatures and approaches, tell it to maintain something I call the papertView more