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 (334)
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,
Get Started on Dev.to! A Beginner's Guide to Engage with the Community! 💡
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! 🚀
Make sure to check out other resources here by other community members on dev.to: dev.to/help/community-resources
Feel free to introduce yourself and welcome others by replying to at least 2 people! It would be greatly appreciated! :D
Hope you are having a great summer and enjoy your stay here on DEV :D
Hey everyone, I'm Amy Lee. Just joined DEV and excited to be here. Looking forward to learning from this community and sharing what I know along the way. Feel free to say hi, would love to connect!
I'll love to connect too; just joined today
"Welcome! Same here, just started posting. What are you working on?"
Hey Amy!
hey
Welcome Amy! What brings you here on DEV specifically?
Hello everyone, I'm new here and very happy to join this developer community.
I mainly focus on backend development, frontend technologies, tool efficiency, and some engineering practices. I hope to learn from everyone's experience here and will try my best to share the problems I encounter, my solutions, and some insights during development.
If any questions are unclear, please feel free to remind me. I hope to communicate and improve together with everyone.
Nice to meet you all!
Hi,Amy
Hey Ami 👋!
Hi Amy! Welcome to DEV! What's your favorite programming language and why?
Hey Amy how are you doing? I'm new to the platform myself.
"Hey Amy, welcome! Just joined too, excited to be here."
hi amy im new too
Hello, Amy!
Hi
Hey Amy !!
Greetings to all developers! 😊
Hey, I'm Alexis, I'm developing things in go!
Hey Alexis, happy to connect !
Hi everyone!
I'm a MERN dev (MongoDB, Express, React, Node), currently learning advance backend engineering. Really excited to be part of this DEV community.
Looking forward to learning from you all, sharing what I'm building, and hopefully making some good connections along the way. Feel free to say hi !
Hello all! Excited to join the community!
Hey Frank, happy to connect !
Hello everyone 👋 , im new here... a little on my bg;
i build multi-agent systems, which mostly means watching very confident software fail in creative ways. agents handing the same task back and forth for days and other sloppy work. fun stuff... i know.
Iam writing a series called "The Harness Layer" , includes depth on my research and analysis on verification, hand-offs/handshakes between agents and servers.. and why evaluation from your model passes but the run still fails... take them as my notes for building towards a robust solution for a community driven RnD project.
anyway thats me boring you, what do you all work on, lets talk!🙌
Agents handing the same task back and forth for days - I fixed that with a counter, not a better prompt. I run a game where AI bots play against each other and against people, and every step advances on a hard message limit enforced outside the agent. Anything inside the agent gets argued with.
What are you using for the handshake between agents - a schema you validate against, or another model reading the output?
yeah makes sense. anything the agent can argue with, it will argue with. hard limit outside the loop is the boring fix that actually works. for the handshake im mostly on the schema side, not another model reading the output. so basically, validate the shape at the boundary, treat prose as the failure mode. a second model as judge just gives you a second opinion,
and yes ive watched that rubber-stamp a bad handoff more than once. the annoying part is when every step passes its own schema and the run is still wrong hence failure is not inside either agent, its between, what i like to call, 'the seams'. thats basically what the harness layer notes are pointing out and hope to bring more research to the table soon...
but curious about your AI bot game... how hardened is your message limit? is it fixed per turn, or does it scale with the game state?
Hey Azan! 👋 I'm new here too and just getting into open source. Your “watching very confident software fail in creative ways” line got me 😂
I'm working on AI-native developer tooling, so your work on multi-agent systems and verification really caught my attention. Keen to check out The Harness Layer!
Appreciate it Max. Fair warning, the series is mostly me documenting ways the tooling loses arguments. Good luck with the AI-native tooling that layer is where half my verification headaches come from anyway.😅
but looking to see if you have anything we interesting we can work on.
Hey Azan — finally read the latest Harness Layer piece properly. The line that stuck: "Generation was nearly free. Verification was the whole job." — and then the way the risk keeps moving, generation → verification → environment, each step less observable than the one before.
I think we're hitting the same problem from opposite ends. I've been writing about what interfaces should expose once agents, not humans, are the operators — one of my core lines is "expose the operational state, hide the implementation." That feels like your point from the other side: if the interface an agent drives doesn't surface its real state, a harness ends up verifying blind.
Which is concrete for you — Orcha plans/routes/verifies across MCP, A2A, and computer-use, so the super-agent is really an agent operating other software through a stack of protocols.
So: what should an interface expose so a harness (or a super-agent built on one) can verify what actually happened without fighting it? Feels like something concrete to build around — would love to compare notes.
Hey everyone! I'm Farzam; full-stack & AI engineer, mostly working on systems that need to actually survive production (not just demo well). Been building web/mobile/cloud stuff for 8+ years, more recently focused on AI pipelines retries, failover, human review gates, that kind of thing. Looking forward to sharing what I learn here 👋
Retries are the part everyone gets wrong first. Mine only work because every step's state is stored, so a failed step reruns on its own, sometimes on a different provider, without replaying the whole thing. Failover sits above the agent, not inside it.
Where do you put the human review gate - before the action, or after it with a rollback?
Before, in this case. The pipeline writes a QA record of open issues and disagreements, then halts right before anything reaches reporting, only the flagged item blocks, the rest of the queue keeps moving. Rollback felt riskier here since a wrong extraction had already fed downstream steps by the time you'd catch it. Curious about your setup though, does rerunning on a different provider ever produce a different-but-still-wrong result, or does that mismatch itself become a signal?
Hi everyone! 👋
We're ArtClick, a web development studio based in Kyoto, Japan. We build company websites, WordPress solutions, and custom web systems for businesses of all sizes.
We're excited to join the DEV community to share frontend development tips, CSS tricks, WordPress insights, and lessons we've learned from building real-world projects. We also look forward to learning from all of you and connecting with fellow developers.
Looking forward to being part of the community! 🚀
Hi everyone!
I'm Luiz, from Brazil. I'm switching careers into tech after a background in Biotechnology and a Master's in Chemical Biology, where I spent years analyzing experimental data and documenting complex processes with a lot of rigor.
I'm currently studying Systems Analysis and Development, learning Python, APIs, and databases, and I'm genuinely excited about applying AI tools to real development workflows (I've been using Claude Code to automate parts of my own job search, actually).
Looking forward to learning from this community and sharing what I build along the way. Happy to connect with anyone else who's also transitioning into dev from a different field!
Hey Luiz — I came into dev from marketing, so I know that jump well. The
"documenting things with rigor" habit helps way more than people expect.
Curious what you automated in the job search with Claude Code?
Thanks Oleksandr! Nothing too fancy yet, but the main thing is a workflow where Claude Code helps me track and triage job postings (matching them against a few "deal-breaker" criteria, like remote/hybrid only), and keeps a log in a private GitHub repo via pull requests, so I have a history of what I applied to and when. Still evolving it, but it's saved me a good amount of manual tracking. What about you, from marketing to dev, did you find anything from that background carried over in a similar "surprisingly useful" way?
Hi Luiz! Nice to see another Brazilian here. I’m also starting my journey in tech, currently studying Computer Science. Your career transition is really interesting. Looking forward to following your progress!
That's interesting career graph, Luiz. Happy to connect !
Hello everyone, I’m Israel. I’m a musician and a CAD designer, and I’m finally entering the tech field—starting out in Systems Analysis and Development. I’ve always admired software development, and although I’m just getting started, I’m optimistic about this future career. I hope to learn and contribute to the community in some way.. 😎
Hey, Israel! Nice to see a fellow musician :). What do you play?
Show, Clarinet and Sax
Oh, hey Israel santos ,are you new here?
Hi Israel! I’m also just starting my journey in tech and studying Computer Science. It’s nice to find other people beginning this path too. Wishing you a great journey!
Hey everyone! 👋 I’m Axis, the Lead Architect at Sunverse AI, based in Abuja, Nigeria. I’m currently building Lawyie, an intelligent legal infrastructure designed to bridge the $10B legal access gap for SMEs and individuals across Africa."
What brought me here:
"I’m deep-diving into LLM orchestration (Llama-3), ultra-low latency inference via Groq, and building robust Python backends. I came to Dev.to to share my journey of 'Building in Public' and to connect with other engineers who are passionate about using AI to solve real-world problems in emerging markets."
Fun Fact:
"I’m a 16-year-old homeschooler! 🏠💻 I recently celebrated a personal milestone: architecting an in-memory PDF generation pipeline that uses SHA-256 cryptographic hashing to 'seal' legal contracts. I love the challenge of turning complex laws into clean, executable code.
Hi everyone — I'm Dean. I write about markets, probability, and the economics around AI: platforms, compute, pricing, incentives. I'm also interested in AI tools as a user: what saves time, what breaks, and what costs more than the demo suggests. Glad to be here.
Hi all, Rickesh here, joining from Singapore.
I work on computer vision and robotics. Nine years of that was autonomy on real vehicles across four countries, and lately I spend most of my time on world models and evaluation for embodied AI.
What actually brought me here: I spent a while convinced a model was bad at reasoning because I had measured it at 0.31. The real number was 0.70. My parser had been discarding its answers on exactly the hardest questions, because the reasoning trace hit the token cap before the answer line was ever emitted. I had already written up the wrong conclusion before I found it.
That became my first two posts, both circling the same theme: measurement failing quietly while every dashboard looks fine.
Mostly here to read and to argue about evaluation. If you work on eval harnesses or robotics data pipelines, say hi.
0.31 to 0.70, and the parser only discarded answers on the hard questions. So the error was biggest where you would trust it most.
I have a daily report on a prod app that quietly stopped counting new users. Eight days broken before I noticed, every number in between wrong, nothing flagged it. There is monitoring on that, and a second loop watching the first loop's errors. Neither caught it. I caught it by not believing a number.
Does anything in your eval harness catch a silent parse failure, or is it still a human doubting a number?
Hi everyone! 👋 We are the Flude Team.
As Documentation Engineers, we joined Dev.to to share our journey of building Flude (Fast Layered Universal Documentation Engine). It is a software product designed to turn raw source code across various languages into high-quality, structured documentation.
We plan to publish articles about our development process, the challenges we face with documentation generation, and how we are solving them.
Looking forward to connecting with other developers and technical writers here. If you're interested in tooling and documentation, feel free to say hi! 🚀
Some comments may only be visible to logged-in visitors. Sign in to view all comments.