Leave a comment below to introduce yourself! You can talk about what brought you here, what you're learning, or just a fun fact about
yourself.Reply to someone's comment, either with a question or just a hello. 👋
Come back next week to greet our new members so you can one day earn our Warm Welcome Badge!
Top comments (111)
Welcome everyone to dev.to! Glad you are here and hope you are well! I am a DEV Community Moderator and my primary goal is to support everyone on dev.to and ensure everyone is having a good time!
To get started, I recommend reading this guide on making the most out of dev.to!
Get Started on Dev.to! A Beginner's Guide to Engage with the Community! 💡
Make sure to check out other resources here: dev.to/help/community-resources
Any questions about DEV and want to get answers from a DEV Community Mod? Leave a comment and come chat here!
Insights on Sloan and the MLH acquisition
Ask a DEV Community Mod! 🚀
Feel free to introduce yourself and welcome others by replying to at least 2 people! It would be greatly appreciated! :D
Hi everyone 😀
A big welcome to all the new members, and welcome back to everyone who is returning. It is so nice to have you all here. I hope you have a great time, make some meaningful connections, learn lots of new things, and enjoy every step of your journey.
Looking forward to seeing you around and chatting with you in the comments. Wishing you all the very best, and I hope you enjoy being part of this wonderful community 💙
He, I'm Raul, a software engineer with a long background in distributed systems. Lately I've been going deep on AI agents: basically taking them apart to understand how they actually work, what they do with tools, context and tokens, and where all of that gets wasted.
Most of my spare time goes into a couple of open-source tools around that: one for running and controlling several agents at once, and one that sits as a transparent proxy between an agent and the model to profile every run and find where you can cut tokens and make it more efficient. Always happy to talk to people working on similar problems.
"Where all of that gets wasted" is the interesting part from where I sit. Token/context waste is usually framed as a cost problem, but a lot of it traces back to interface decisions upstream, like re-sending full conversation history because the UI has no cheaper way to remind the model what's already been established. Would be curious what your profiling turns up as the biggest single waste source, tool calls, context re-sends, or something else entirely.
At the moment I am focusing on tools. I am still in a phase of disbelief because I saved between 60 and 80% the cost on tokens, with a simple example of around 50 calls (the AI has to find and fix some tests). I will try to improve the benchmark tonight.
Hi all, I'm Daniel, a solo developer working on Revue, a multi-agent code review tool that runs locally in Claude Code before you commit and again in CI on pull requests. I've been building it since March and testing it on its own codebase throughout, which has forced me to fix every rough edge pretty quickly. I just published my first post here about how AI-assisted coding shifted code review from a minor step to the real bottleneck. I'll probably keep writing about that: code review, AI-assisted development, and the unglamorous side of shipping alone. Good to be here, and curious what everyone else is working on.
Hey everyone 👋 I'm Omar full-stack engineer, currently doing a CS master's, and building in the AI/dev-tools space.
What brought me here: I've been shipping a lot lately and want to be around other people who build. Right now I'm working on Clanker Support an embeddable AI support agent with one core idea: it should answer from your docs, then hand off to a human the moment it can't (with the full conversation), instead of trapping people in a chatbot loop. It's open-source and self-hostable, which is partly why I like the DEV crowd.
Currently learning a lot about the unglamorous side of shipping rate limiting, escalation state machines, and how many ways a "simple" widget can break in production 😅
Happy to be here will go say hi to a few folks below.
The "hand off to a human the moment it can't, with the full conversation" line is the part most support bots get wrong, they either never escalate or escalate with zero context and the user has to repeat everything. That handoff moment is honestly a UX problem as much as an engineering one. Curious how you're deciding the threshold for "can't answer", is it confidence-based or more rule-driven right now?
Totally agree it's a UX problem wearing an engineering costume the mechanics of the handoff are easy, it's the judgment of when that's hard.
Right now it's deliberately more rule-driven than confidence-based, and honestly that was a considered choice. LLM "confidence" (token probs, self-reported certainty) turned out to be a bad escalation signal models are confidently wrong all the time, so thresholding on it either escalates too much or trusts a hallucination. So the primary triggers are explicit: the visitor asking for a human (highest priority if they ask, they escalate, no arguing with them), and the agent hitting the edge of its retrieved context (no relevant source -> it says so and offers a human rather than filling the gap).
The interesting failure mode we hit: the agent would keep "trying" after a human had already joined the conversation so we had to build the stop condition as carefully as the start one. Confidence-scoring is on the roadmap as a secondary signal, but I'm wary of making a probabilistic number the thing standing between a frustrated customer and a human. Curious if you've seen confidence-based escalation actually work well anywhere my prior is that it's seductive but unreliable.
Hey all, I'm Rohit, a Lead UX Designer working in core banking and fintech. I've been writing here about UX for AI products, agentic interfaces, and error-state design, basically what happens to an interface when an AI feature fails, hesitates, or only gets part of the job done, since that's a whole category most design systems weren't built for. Looking forward to trading notes with people building in the AI/agent space.
Hello! I'm Haya, a backend engineer from Saudi Arabia with a strong interest in the fintech industry. I joined DEV to grow my knowledge as a backend engineer and to share the projects I'm building in my free time. I'm excited to connect with this community, learn from others, and contribute what I can along the way.
Outside of work and engineering, I'm passionate about music and sports.
Hi everyone! 👋
I'm Malik Taleeb Shahbaz from Pakistan, a Frontend Developer and IT student. I enjoy building fast, responsive, and SEO-friendly web applications using Next.js, React, and TypeScript.
I joined DEV to learn from the community, share my projects, and write about web development, performance optimization, and modern frontend technologies.
Looking forward to connecting with you all!
Hello to all, I am Abdul Imran from India, a golang leaning developer who make cli and tui tools .
i have around 1+yr of experience and joined today to tell everyone about my projects.
I learn and grow , and i am curious to work with other developer who can catch up to me and make one of the best utility tool with me.
thats it , hope you like it :)
Hey everyone! I'm Marvin, solo founder from Paris/Lisbon. 15 years in retail and e-commerce.
Currently building four things:
I'll be posting about the Astro multilingual architecture and the WhatsApp-to-book automation pipeline. Happy to connect with anyone building in e-commerce or no-code automation!
Hey DEV community! 👋
I'm building KollabKit - an AI-powered media kit builder for Instagram creators. Connect your Instagram, get a live shareable media kit in under 30 seconds. No PDFs, no Canva, no manual data entry.
We just launched and currently have a media kit builder and rate calculator live. Building more like caption suggestor, AI video draft approver for brands, creator discovery.
Would love feedback from this community on the product and what we're building. Happy to answer any questions!
Some comments may only be visible to logged-in visitors. Sign in to view all comments.