Community feed

I’m building an AI voice assistant — what features would actually make you use one daily?

I’ve been working on a personal AI assistant that can have conversations, help with planning, answer questions, assist with studying, and handle everyday tasks. I’m curious about what developers and tech users actually want from an AI assistant beyond just chatting. If you could have one AI assistant feature that genuinely saved you time every day, what would it be?

What’s the best way to automate driving licence data entry?

One thing we noticed while working on document automation is that onboarding can become unnecessarily repetitive when customers have to manually enter information from their driving licence. A driving licence already contains useful information such as the name, address, licence number, and expiration date. The challenge is getting that information from an uploaded image into the application without making someone type everything manually. This is one of the problems we’ve been working on with our Driving Licence OCR API. The basic workflow is quite straightforward: Upload licence → Extract information → Use the data in your application The interesting part is what happens after extraction. OCR isn't the same as verification. OCR handles converting the document into machine-readable data, while verification can be handled as a separate step depending on the application's requirements. For us, this approach makes sense for use cases such as car rentals, insurance, banking, customer onboarding, and digital record management. I'm curious how other developers approach this: Do you build document OCR internally, use a third-party API, or combine OCR with your existing verification workflow? Would be interesting to hear what has worked well for others. Source: AZAPI Driving Licence OCR API

Has anyone's company actually banned or restricted AI coding tools, and why?

Heard of a few companies restricting tools like Copilot/Cursor over IP or security concerns. Curious about real examples and what specifically drove that decision.

Async-first teams of 2026 — what replaced your daily standup?

We went fully async last year and dropped the daily standup for a written update in a shared doc. Works well for us but curious what other distributed teams are doing differently, especially across IST/PST time zones.

Is contributing to open source still a good use of time for career growth, or has AI-generated PRs diluted its value?

Used to be a strong signal on a resume. With a rise in low-effort AI-generated PRs flooding some repos, wondering if maintainers and recruiters still weigh this the same way.

Is a machine learning specialization still the safest bet for new grads, or is the market oversaturated already?

Every third student in my batch is pivoting to ML/AI specializations. Wondering if this is going to lead to an oversupply the same way full-stack bootcamp grads flooded the market a few years back.

What I learned building my first real Make.com project

Spent the last few weeks working hands-on with Make.com inside AI Automation. Sharing a few practical lessons that would have saved me time if I'd known them on day one.

What I learned building my first real Gemini project

Spent the last few weeks working hands-on with Gemini inside AI Tools & Platforms. Sharing a few practical lessons that would have saved me time if I'd known them on day one.

What's the best way to get started with Design Systems in UI/UX Design?

Trying to build a solid foundation in Design Systems as part of UI/UX Design. Any recommended resources, or a real project idea to practice with, that actually reflects how it's used in production?

What I learned building my first real Identity & Access Management project

Spent the last few weeks working hands-on with Identity & Access Management inside Cybersecurity. Sharing a few practical lessons that would have saved me time if I'd known them on day one.

Rust for backend services in 2026 — hype or genuinely worth the learning curve for a startup?

Considering Rust for a new high-throughput service instead of our usual Go/Node stack. Team has zero Rust experience currently. Is the performance gain worth the ramp-up time for a small team?

How we evaluate a new SaaS vendor before signing — our internal checklist

After getting burned twice by tools that looked great in the demo but failed on data export and support SLAs, we built a 12-point vendor checklist covering security certifications, data portability, uptime history, and exit clauses. Sharing it here in case it helps others avoid the same mistakes.

What I learned building my first real Spring Boot project

Spent the last few weeks working hands-on with Spring Boot inside Software Development. Sharing a few practical lessons that would have saved me time if I'd known them on day one.

Nice to join the community!

Should junior developers be worried, or is this cycle similar to past disruptions like the shift to cloud or mobile?

Older engineers in my company keep comparing this AI moment to the cloud transition, saying it created more jobs than it destroyed. Is that comparison actually holding up so far, or is this genuinely different?

What a year of running our own hardware actually cost us

We moved our staging environment onto a refurbished server in the office to cut cloud spend. Twelve months on, here is the honest accounting. Saved: roughly ₹31,000 a month in cloud bills. Spent: ₹92,000 on the machine, ₹18,000 on a UPS after the second power cut corrupted a disk, two full weekends of setup, and about one evening a month on maintenance. It broke even around month five. Whether it was worth it depends on how you value the weekends. What genuinely surprised me: the reliability was fine. What did not surprise me: nobody else on the team wanted to learn how to look after it, so the bus factor is one — and that one is me. Would I do it again for staging? Yes. For production? Not without a second person who knows the setup.

Is anyone actually shipping AI features, or are we all still prototyping?

Every team I talk to has an LLM prototype. Very few have one in front of paying users. The gap seems to come down to unglamorous things: • What happens when the model returns nonsense at 2am and nobody notices? • How do you test something whose output changes between runs? • Who pays when a single user pastes a 40,000-token document? I am not asking whether it is useful — clearly it can be. I am asking what it took to get from demo to production, and what you cut to get there. If you have shipped something real, I would genuinely like to hear what broke first.

A backup you have never restored is not a backup

We had nightly database dumps running for two years. Green ticks in the dashboard, files in object storage, everything looked fine. Then we needed one. The dumps were being written with a flag that excluded large object data, and nobody had ever tried loading one into an empty database to check. What we do now, and I would recommend to anyone: • A scheduled job restores last night's dump into a throwaway database. • It counts rows per table and compares them against the source. • It runs one real query the application uses, and fails loudly if the result is wrong. It takes about four minutes a night. The first time we ran it, it failed — which is exactly the point. If you take one thing from this: the question is not "do we have backups". It is "when did we last restore one".

How are you handling secrets rotation without a paid vault?

Small team, Postgres + a Node API, everything on a single cloud provider. Right now our database password and third-party API keys live in environment variables and have not been rotated since we set them up. Managed vaults are the obvious answer but the pricing starts above what this project can justify. What I am considering: 1. The cloud provider's own secret manager — cheap, but ties us to them harder than I would like. 2. SOPS with age keys committed to the repo, decrypted at deploy. 3. Just rotating manually on a calendar reminder and accepting the risk. For those of you running small production systems: what actually works in practice, not in theory? I would rather hear "we do option 3 and it is fine" than a best-practice answer nobody follows.

We moved 12 microservices off Kubernetes and our bill dropped 60%

We ran a 3-node managed cluster for eighteen months because that is what everyone said to do. Looking at the actual numbers, eleven of our twelve services handled under 5 requests per second at peak. What we changed: • The two services that genuinely need to scale stayed on Kubernetes. • Everything else moved to a pair of load-balanced VMs behind a process manager. • Cron jobs became actual cron, not CronJob objects. Monthly infra went from around ₹58,000 to ₹23,000. Deploys got slower to set up but faster to reason about — no more debugging why a pod was Pending at 2am. The honest trade-off: we gave up automatic bin-packing and rolling deploys with no downtime. For a team of four, that was worth it. If you are twenty engineers with fifty services, ignore all of this. Curious whether anyone here has gone the other way and regretted it?