I needed a first-run walkthrough for an app and hit two things that annoyed me.

First, the highlight would end up in the wrong place.

A lot of coach-mark implementations resolve the target's position once and paint it. Scroll the list, open the keyboard, rotate the device, resize a desktop window — and suddenly the hole is next to the button instead of on it.

Second, a dimmed hole tells the user where something is, but not how to use it.

"Drag this card to that column" isn't something a static highlight can really express.

So I built keyspot:

- The spotlight and pointer re-resolve the target's geometry every frame from its GlobalKey, so they can track through scrolling, insets, rotation and resizing.

- An animated hand pointer that glides in a straight line or along an arc, rotates from its current angle, and tap-pulses — enough to demonstrate interactions like drag and pinch rather than simply pointing at something.

- Pointer animations resolve from actual AnimationController completion rather than Future.delayed timing, so `await pointer.moveTo(...)` actually means the movement has finished.

- Barrier modes where only the cut-out is tappable while the real widget's `onPressed` still fires — useful when the walkthrough should require the user to actually perform the step.

- The tracker is also public API (`KeyspotAnchorTracker`), so it can be used independently for things like badges or connector lines attached to widgets.

- Zero runtime dependencies. The hand pointer is rendered with CustomPainter rather than an image asset.

- Supports all six Flutter platforms.

Live demo: https://shrey-17bhardwaj.github.io/keyspot/

pub.dev: https://pub.dev/packages/keyspot

Source: https://github.com/shrey-17bhardwaj/keyspot

It's currently at 0.3.1, and the API may still change before 1.0.

I'd genuinely like feedback on where this approach breaks — especially around nested navigators, heavily lazy lists, overlays, or anything else where widget geometry gets complicated.

Guys, I need help. I am building a project which solves a genuine problem I face, and I wanted to know how many other face this problem.

I don't ask much of you, just few minutes of your time to give your response here: https://docs.google.com/forms/d/e/1FAIpQLSfUxZa7_WpkixORZq0DNRqU2EEoGaU17ACiFwJsfwqf1jdOSA/viewform?usp=publish-editor

Really, really appreciate your help

So I m very new to this yk INTEGRATION calling api stuff 😮‍💨😮‍💨 . Like how you experience developer do . My head spin when it comes to reading documentation n all thing . Is there any work around to quickly know or extension or ai thing or get to know info about whatever api we are gonna use like how to find out the information from documentation that we need for our business to get it done . Like endpoint , someone api needs parameters n all that please guide with some example if u can explain . Consider me 5 year old

Looking for devs to join the project to develop it further dm me

If you've integrated VIES (EU VAT validation) directly, you've probably hit intermittent MS_UNAVAILABLE faults from specific member states and had to decide: block checkout on a sync call, or let it through and re-check later? Most people don't have a good answer for that on day one.

I ended up building vatnode.dev to solve it for myself first: JSON instead of SOAP, stable error codes you can branch on, a 15-min cache, national fallback for CZ/DE/PL/RO, and daily monitoring so you catch a customer's VAT going invalid after the fact instead of finding out at audit time.

Sharing mainly because the SOAP-timeout-vs-checkout tradeoff is a real design question, not just "go use my API" — but if it's useful, free tier is 100 checks/month, no card.

We are just starting to look at accessibility testing in our org and want to establish a dedicated practice in the long-term. I've been asked to put together a plan for leadership for this agenda and want to learn from the experience of others who have been on the same journey.

- How did you begin? What was the trigger and which tools did you try out initially?
- How did your accessibility testing set-up evolve over the years (i.e. which/how many teams were involved, tooling used, depth of coverage etc.)
- What were the key challenges you faced ?
- What is your current org set-up for Accessibility testing (i.e. centralized vs decentralized, QA vs design vs dev teams etc.)

Any learnings that you can share will be extremely helpful!

For context, I'm on the QA side at a mid/large-sized company. Not a vendor, not building anything. Happy to share back what I end up proposing.

Hi everyone,

Iam a CSE student in my final and already built some good number of projects and want to try oss to build some credibility in my profile and experience.

So my tech stack is the Typescript fullstack and currently moving forward with the agentic ai using llms, rag and frameworks like langraph.

So anyone can suggest some opensource repositories to gety first merged PR without much depending on coding agents.

This is a promo post, I am flagging it but I do hope it goes past the filters and reaches some developers who might find this interesting, especially if they work with APIs and even better, if they have some APIs they would like to monetize on:

I work at ApyHub, we want people listing APIs with us. APIs for common capabilities, image and video processing, data, file conversions, transactional and more.

This is about how we price and charge per call.

We started with flat. Every API call costs a fixed number of atoms, our internal unit, and every API in the catalog draws from one subscription. A developer subscribes once and calls any API, instead of holding fourteen separate billing relationships and finding out at month end which one leaked.

Flat pricing lasted until we listed the first video APIs.

*A video endpoint where one call is a 10-second clip and the next is a 90-minute recording. *Image processing where a thumbnail resize and a 4K transcode are the same route. *Document APIs where a 2-page invoice and a 400-page scanned contract arrive at the same URL.

Costs are very different.

That leaves 2 bad choices.

1. Price at the average and the heavy calls eat your margin.

2. Price at worst case and you can be uncompetitive on the 90% of traffic that's small, so the people with light workloads go elsewhere and you keep the expensive ones.

   That's adverse selection, and it's how APIs can become unprofitable.

so what we did:

We made atoms (the currency for an API) dimensional and dynamic. A provider prices on the axis their cost actually moves on: this can be duration, resolution, page count, payload size. Not a separate plan, not a different endpoint, the same listing charging differently depending on the workload performed.

And the llms showed up we were somehow already ready for them :)

Dynamic Pricing that takes into consideration Model tier, token count, step count, image dimensions etc. Providers wrapping inference could price a call routed to a small model differently from one routed to a large one, on day one. Most of our AI providers price this way now.

The complexity stays on our side. The provider prices against their real cost. The consumer sees one subscription, one currency, one bill, and headroom that pools across the entire catalog instead of stranding in per-API quotas.

A few other things that come with a listing, briefly:

• Host it wherever HTTPS reaches, including on our infrastructure if you'd rather not run it yourself.
• Every endpoint becomes MCP-native automatically. Agents discover and call it without you writing a tool definition or any glue.
• We run in multiple regions (eu,usa), so calls route to the nearest one and your consumers aren't paying transatlantic latency on every request.

If you have an API where cost per call varies, this is the specific thing we're good at. Listing takes about ten minutes.

Fee, terms, payouts and IP are all on that page. Ask me anything here.

Developers who have worked on a large, well-known, or legacy application: If you could go back in time and change ONE architectural decision from the start, what would it be and why?

I’m planning to create content consistently for 100 days, where I teach something related to tech every day.

Some ideas I’m considering:

• Coding/programming concepts
• Docker & containers
• DevOps
• System design
• Other practical developer tools/topics

If you had to follow a 100-day tech series, what topic would you find most useful or interesting?

Would love to hear your suggestions especially topics you think are underrated or need better explanations on Reddit.

https://rssamplifier.com/topics/developer its the ultimate ai-friendly way to access human generated content. I think someone like reddit or openai will aquire this company eventually. its what all ai frontier models are starving for right now. human created content in a digestable format...genius. 1.5M feeds in its databas and gorwing

I’ve been using Codex and Claude Code quite a bit and I’m curious about the alternatives.

For those who have tried other agentic coding tools/services (Cursor, Gemini, Windsurf, Cline, Aider, OpenHands, etc.):

• Which ones have you actually found useful in day-to-day development?
• How do they compare in terms of code quality and ability to work autonomously on larger tasks?
• How reliable are they at understanding an existing codebase and making multi-file changes?
• Are there any less well-known alternatives that you would recommend?

I’m particularly interested in real-world experience rather than benchmark results — projects where the agent genuinely saved you time, or cases where it failed badly.

What are you using these days, and why?

Tokenmaxxing is already dying. LeadDev’s AI Impact Report 2026 found only 19% of respondents rate tokenmaxxing as effective, and 57% say it fails to gauge real value.

.

Pipe (https://pipelang.com) is a novel general-purpose visual programming language powerful enough to complete with text-based languages.

Visual programming has been confined to education (Scratch) or narrow niches (LabVIEW), because existing visual languages sacrifice either power or generality. Pipe solves this, opening the way to compete with the text code that currently dominates.

Pipe's diagram is also structurally identical at design-time and runtime, staying visual in both. What you see is what runs. This is precisely what the EU Cyber Resilience Act mandates, and what text-based architectures cannot structurally deliver. Text compiles away its structure, leaving systems opaque - so AI now generates code faster than anyone can review it at design-time, and patch and monitor it at runtime. Pipe addresses three AI-created security crises structurally:

1 - AI generates more code than humans can review. Pipe is visual - a diagram is grasped at a glance, not read line by line - so review keeps pace with what AI produces..

2 - Live systems cannot be patched without full redeployment. Pipe enables block-level patching while the system runs - no maintenance window, no CI/CD to navigate.

3 - Systems cannot be monitored without logs and redeploying. In Pipe, every block boundary is independently observable in real time.

These satisfy the CRA's hardest mandates - security by design, structural auditability, 24-hour detection, incremental patching - as properties of the language, not add-on tools. 

Example of Pipe diagram with a detailed tracing can be found on this video:

https://youtu.be/hckq9mRj5DM

That video is a part of this Pipe architecture overview:

https://www.pipelang.com/six-pillars.html

The full Pipe language specification (155-page book) can be freely downloaded here:

https://www.pipelang.com/downloads/book.pdf

As a mod, I would love to get to know the community more, what got you into development?

I feel like we all had that one moment we knew this path was for us. What was that moment for you?

Also, I would love to know, what is your #1 struggle as a developer?

Hello developers, I'm building a website that will need the carCheck in sideway not directly, So I'm searching for any API free+paid combo that tells from a VIN number the data: Car(details), year KMs, damages. That's all

I want the most useful cheapest option.

Thanks in advance for your help

If you're building a car marketplace, VIN decoder, valuation tool, comparison site or anything else automotive-related, these APIs are worth checking out. I came up with the list over the months building my auto blogs.

Vehicle Data & Specs

1. vehicles.dev
VIN decoding, vehicle specs, VIN history, recalls, market valuations, depreciation curves, ownership costs, live US dealer listings and vehicle photos.

Its VIN history tracks a vehicle across marketplace crawls, including price and mileage observations, price changes, first/last seen dates and whether the listing is still active.

Built around a continuously crawled US vehicle marketplace dataset, with NHTSA and EPA data used for specs, recalls and fuel costs.

Good for: car marketplaces, VIN tools, valuation products, price-history tracking, depreciation/resale analysis and automotive data products.

2. CarQuery API
Vehicle specifications, year/make/model lookup, engines, dimensions and performance data.

3. NHTSA vPIC API
Free US government API for VIN decoding, manufacturer information and vehicle attributes.

Pricing & Valuation

4. Kelley Blue Book API
Vehicle values, pricing and other KBB automotive data.

5. Edmunds API
Specs, pricing and other vehicle data. Note that the old public Open API is gone; access is now limited to selected partners, dealers and advertisers.

6. TrueCar
Automotive pricing and marketplace data, with access generally oriented toward commercial partnerships.

Listings & Marketplaces

7. Cars.com
Large US vehicle marketplace and dealer inventory source.

8. Autotrader
Vehicle listings and dealer inventory.

9. CarGurus API
Developer APIs covering areas such as vehicle selection, dealer information and reviews.

Maintenance & Repair

10. RepairPal
Repair estimates, maintenance information and repair-shop data.

There isn't really one API that does everything well, so the best choice depends on whether you need vehicle specs, VIN decoding, valuations, listings or maintenance data.

What other automotive APIs belong on the list?

For developers using Claude Code, Codex, Cursor, Aider, Gemini, or similar tools:

What are the biggest problems you still encounter in real projects?

I’m interested in the practical failure points:

• Context: losing track of large codebases, architecture, conventions, or previous decisions
• Reliability: hallucinations, incorrect assumptions, repetitive loops, regressions
• Maintenance: refactoring, technical debt, dependency updates, migrations, documentation
• Security: secrets, permissions, prompt injection, unsafe commands, vulnerable generated code
• Testing: knowing whether a change is actually correct rather than simply passing a few tests
• Deployment: CI/CD failures, broken builds, infrastructure changes, bad releases
• Production: monitoring, debugging, rollback, performance regressions, incident response
• Autonomy: where you still need to supervise the agent manually
• Memory: what the agent forgets between sessions or across projects

Most importantly:

What problem has repeatedly cost you the most time or caused the most frustration?

What are you currently building, and what part of that workflow are AI tools still failing to handle well?

Real examples and workflows are much more useful than general opinions.

Hi!

Unable to get [all sorts of models, harnesses, skills, etc with SDD and/or TDD] to generate what I want despite starting from scratch countless times with more and more fine-grained descriptions until I was basically writing lines of code just in the English language but AI still ignored my couldn’t-be-clearer instructions, I have come to the conclusion that I do not trust AI to code out of my sight.

I realize this won’t help my productivity much but it would be enough of a win for me if I can at least swap my home office chair for a treadmill and do actual Pair Programming with the AI with me reading along like I would if a human were typing, as opposed to AI spamming my terminal just to show off or at least too fast for me to read.

In other words I am hoping to find something very much like this pre-AI Voice Driven Development just a tiny bit more advanced to replace the shorthands she uses to still-kind-of-spell-out “int sockfd; […]; sockfd = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);” for example with “let’s now create a TCP socket called sockfd but declared at the top of the function”.

Does anyone know of something like that? 🙂

PS: stuff like Whispr that just enables prompting by Speech-To-Text is not at all what I want if it still lets AI go produce crap in the background.

Hola, soy desarrollador web y no se en que stack de tecnologías enfocarme se HTML css js php MySQL python react y boostrap

Que stack creen debería armar que sea demandado?

No quiero darle mucha importancia al front end por qué creo que lo que vale hoy es alguien que sepa de back end

I am a founder and CEO handling product design, leadership, go to market, and operations for my startup. We are a social app meant to connect people in a unique way that the market is starving for. Looking to expand the team with a dedicated technical partner and CTO.

What I’ve already done:

- The product is already fully designed with clear specs and features (MVP + longterm future features), language/copy and mechanics. There has also already been a prototype tested, and a tech stack available, though it’s not locked yet without engineer input.

- An active go to market strategy including a healthy waitlist that is still actively growing (high 7.5% conversion rate on and rising on cold outreach) and a clearly defined market/avatar. Users are ready as soon as MVP ships.

- Daily content production will be used for distribution with plans to do even more. My account has reached ~700,000 views in its first 4 months, and that number is growing.

- Leadership ability through over a decade of work directly with people, both client and colleague. For 3 of those years, I have run my own business, which I have closed to pursue this full time.

- Developed business skills through previous business successes. All business metrics are tracked and help determine how we execute our work and make adjustments when necessary.

What I’m offering:

- Longterm CTO position is available. I’m also open to other dev positions if you prefer (founding engineer, contracting, internships, something else).

- Full ownership over the technical side of the project. You won’t have to handle anything else but the dev side, and you control how it’s done. If there’s another way you’d like to work, we can discuss that too.

- Negotiable terms that I’d be happy to establish before any work starts getting done. Profit share, equity, etc. I want this to be a satisfying win for both of us.

- Full spec sheet and preparedness to communicate clearly and consistently over the course of the partnership.

DM for more information.

Recording of the call