DEV Community

Cover image for Welcome Thread - v399

Welcome Thread - v399

Sloan the DEV Moderator on October 28, 2026

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. Re...

Displaying a subset of the total comments. Please sign in to view all comments on this post.

Collapse
 
francistrdev profile image
FrancisTRᴅᴇᴠ •

Welcome everyone to DEV! We are glad you decide to join us!

As mention before for many welcome threads before me, I will make my introduction at Welcome Thread v400! I made a promise to @javz that I will do this (though forgot the reason why lol).

In the meantime, feel free to visit my guide on dev.to!

I will see you all next week there! In the meantime, what brings you to DEV?

Collapse
 
javz profile image
Julien Avezou • • Edited

Looking forward to your intro @francistrdev !
It's almost time already! 🍿

Collapse
 
francistrdev profile image
FrancisTRᴅᴇᴠ •

I'm debating if I should make it short and sweet or a whole research article length lol

Thread Thread
 
paresh_chaudhari_14ca7efb profile image
Paresh Chaudhari •

Hey! 👋 Nice to connect with you. What technologies or projects are you currently working on? Always great to learn from fellow developers!

Thread Thread
 
francistrdev profile image
FrancisTRᴅᴇᴠ •

Hey Paresh! Just working on Next.js mostly. What about you?

Collapse
 
surfasahthi profile image
Zeenathul Surfa Sahthi •

Thank you for the warm welcome!

I'm excited to be part of the DEV community and looking forward to learning, sharing my experiences, and connecting with developers from around the world.

I'll definitely check out your guide. Looking forward to your introduction in Welcome Thread v400!

Collapse
 
ben profile image
Ben Halpern •

Welcome welcome everyone!

Collapse
 
baigorria profile image
Guillermo Baigorria •

Hi, I'm Guillermo, Head of Technology at Werk. We're building an AI assistant that keeps the work of teams on the move organized from their voice notes and photos. I'll be writing about model-agnostic harnesses, LLM routing and testing agents. Happy to be here.

Collapse
 
postking_app profile image
Dite Gashi •

Welcome Guillermo! LLM routing is very relevant to me. I run most of PostKing's generation through OpenRouter and send Claude calls direct. The biggest lesson so far: the same prompt doesn't behave the same across models. One model left reasoning on by default and quietly cut our blog posts short; another sometimes returns stray string fragments inside a JSON array. Do your harnesses validate each item, or do you retry the whole call? Following for the testing posts.

Collapse
 
deepanjan_prasad profile image
DEEPANJAN PRASAD •

Hello

Collapse
 
devamahesh198 profile image
devamahesh198 •

Hello! Your work sounds really cool!

Collapse
 
divyansh_tripathi profile image
Divyansh Tripathi •

Hey @francistrdev 👋

I'm Divyansh, and I'm exploring how technology works behind the scenes, especially in the world of AI and software engineering.

Lately, I've been spending a lot of time learning about LLMs, RAG, AI agents, embeddings, and how these technologies are being used to solve real-world problems.

I enjoy breaking down complex technical concepts into simple, practical explanations that are easy to understand, whether you're a beginner or an experienced developer.

I'm here to learn from the community, share what I discover, exchange ideas, and connect with people who are equally curious about technology.

Looking forward to reading what everyone is building and learning here.

Collapse
 
thomasbnt profile image
Thomas Bonnet •

Welcome to news members on DEV!

Collapse
 
muntazer_mahdi profile image
Muntazer Mahdi •

thank you!

Collapse
 
sachintatrari profile image
Sachin Tatrari •

Thank You Thomas.

Collapse
 
shubhradev profile image
Shubhra Pokhariya •

Welcome to DEV! 😊

One of the things I love about this community is how it makes you feel that you're part of something. People take the time to read your work, encourage you, share their own experiences, help each other, and celebrate what you create. It's such a good feeling to see your work reach people and get appreciated, and to know that something you've shared might help someone else too. That's something I really value about DEV.

I'm so glad this community brings developers together like this, and I hope all the new members feel that same warmth here too. It's lovely to have you all with us. Wishing you all a great time being part of DEV! ☺️

Collapse
 
nehamangla profile image
Neha Mangla •

Hey everyone!

I'm Neha, a software engineer with 7+ years of experience, mainly working with React, TypeScript, and Node.js.

I enjoy working on complex web applications, especially frontend architecture, state management, debugging, and performance.

I've just joined DEV and published my first article, "Why Building a Reliable Calendar UI Is Harder Than It Looks." It shares a few lessons from working with scheduling interfaces, date handling, and frontend reliability.

If you're interested, here's the article:
dev.to/nehamangla/why-building-a-r...

Happy to hear how others approach similar challenges. Looking forward to connecting with the community!

Collapse
 
postking_app profile image
Dite Gashi •

Welcome Neha! A calendar post caught my eye: I build a content scheduler, and timezones plus "what day is this post actually on" have produced more bugs than anything else. Storing everything in UTC fixed the data but not the UI edge cases. Reading your article now.

Collapse
 
fm profile image
Fayaz • • Edited

Welcome all the new comers to DEV. 👋

It's a great community of developers, for the developers.
It's one of only a few places left where conflicting developer opinions rest in peace 🤣🥰

In case you are interested, here's one of my last posts that you may find worth reading:

Collapse
 
a2clab profile image
Addtocalendarlab •

Hi all 👋 We're Add to Calendar Lab. We compare add-to-calendar tools, the services behind those "Add to calendar" buttons on event pages and emails.

We came here to write about what we've learned building it: why calendar buttons break in Outlook, the .ics file rules nobody follows, and how we made the rankings checkable (the scoring code is public, so you can rerun it with your own weights).

Fun fact: Yahoo Calendar's add-event link ignores the Z on a UTC time, so the same link shows a different hour depending on where you open it. There's no official documentation for any of these links, so we keep finding things like that.

Glad to be here. If you've ever fought with time zones, say hi.

Collapse
 
ahmad_shawar_de8d35d546cf profile image
Ahmad Shawar •

Hi everyone! I'm Ahmad, a senior software engineer with 14+ years building SaaS platforms end to end. These days I'm focused on AI agents, LLM retrieval and the systems that make them reliable. I just published my first post here, about why markdown plans break when you run multiple AI agents at once, and I'd love to hear how others coordinate their agents. Glad to be here!

Collapse
 
dsiacci profile image
Dominique Siacci •

Hi Ahmad and welcome ! A shared markdown plan broke the same way here: two workers could take the same unit. What held was a queue they pull from, with one session allowed to merge, and a dashboard so a stuck worker parks the question and takes the next unit instead of leaving the item marked in progress. I'd watch expiry on those claims: without it, a dead session still owns the unit.

Collapse
 
ahmad_shawar_de8d35d546cf profile image
Ahmad Shawar •

Thanks Dominique! That's almost exactly where I ended up. Workers claim tasks from Postgres with a TTL, so a dead session loses the claim when it expires and someone else picks it up. I also added limits on refreshing a claim, because a stuck agent can keep refreshing forever without making progress.

Collapse
 
postking_app profile image
Dite Gashi •

Welcome Ahmad, and congrats on the first post! The markdown-plans problem sounds familiar. I run several coding agents in parallel on one repo, and what helped most was giving each one its own git worktree plus one shared lock for the heavy checks, so they stop stepping on each other's files and RAM. Shared plan files are still the messy part. Off to read your take.

Collapse
 
martinkr profile image
Martin Krause • • Edited

Hi everyone!

Welcome to the dev.to community!

Please tell me in the comments: What brings you to DEV?


I'm Martin, and I worked as a Senior Front-End Architect and Engineering Manager at some of the world's largest digital agencies, where I led front-end teams on international projects. After 15 years, I left.

Now, I travel the world as a professional scuba diver and work as a contracted Tech Lead, Solution Architect, and Technical Recruiter for international clients. Additionally, I write about technology on dev.to and Golem (German).


A life, Jimmy, you know what that is? It's the shit that happens while you're waiting for moments that never come.

That is my advice for everyone. Jump right in. Start writing! Today! That's how I became a published author for No Stach Press and Packt Publishing - who knows what's waiting for you?

My latest book Claude Code Beyond the Prompt will be available in November. Over the next few weeks, I'll run a series on dev. to revealing the chapters one-by-one. Take a look :)

If I have enough time, I write more in-depth articles and thought pieces here on dev.to, for example: Toward Agent Native Development or the follow-up Why Anthropic’s Scion Forces Us to Think "Agent-Native"

My first book, The Complete Developer, is an easy, hands-on introduction to building web applications with Next.js, MongoDB, and Mongoose. It is part of multiple university syllabi in the USA and India - you can get it from No Starch Press


I recommend following the general tags webdev, ai , career and productivity and javascript, react, CSS. Head over to your Dashboard and adjust the tag weight to personalize your home feed.


Here's a list of articles to start your journey as a successful developer and a great writer:


If you are looking for your next job, I highly recommend buying the Book De-Coding the Technical Interview Process by Emma Bostian. It is hands down the best preparation for every technical interview.


If you're interested in my articles, have questions, or want to get in touch, don't hesitate to contact me on dev.to/martinkr

And remember to start posting today, maybe just by telling us in the comments: what brings you to DEV?

Cheers!

Collapse
 
retrytax profile image
Vlad wrinfo •

Hi. My small obsession is the cost of a finished task rather than a finished call: retries, cache, human review, all of it in the same place. Mostly Python and log parsing.

Curious what people here count as "done" when they evaluate an agent run

Collapse
 
connabyss profile image
CH •

Great question. My answer so far: a run is done when the thing it says it fixed is actually fixed, checked by a test that fails without its change. A lot of "done" I've seen only means the tests passed, which isn't the same thing. Do you count human review time in the cost even when the review finds nothing?

Collapse
 
retrytax profile image
Vlad wrinfo •

Counted either way. A review that finds nothing costs the same hour as one that finds something, and skipping the empty ones just makes my average look better than the workflow is. I want the number over every review, not only the ones that caught something

Thread Thread
 
connabyss profile image
CH •

Makes sense. Same with the checks I run on AI commits: most come back clean, and those clean runs are the real price of catching the few that aren't.

Collapse
 
jesseships profile image
Jesse Griffin •

Welcome! If writing feels hard at first, start with a short "what I learned this week" post. It's low pressure, and it creates the habit of finishing things.

Also, hang around the tags you actually use. Being a consistent commenter early on is the easiest way to be part of the community rather than posting into the void.

Collapse
 
camerondev profile image
Cameron Bennett •

Hi everyone. Something that helped me: write the post you wanted to find when you were stuck. Not the general overview — the specific thing, with the exact error and the version. I've written maybe five posts like that and they consistently get the most replies, because someone else is stuck on the same thing right now.

Collapse
 
fisher_38 profile image
Skyler Fisher • • Edited

Hi everyone, full-stack person on Rails and Postgres. Advice I keep giving: cut your first post down to one screen — one bug, one query, one diff. The posts I still revisit are all that size, and they're easier to update later when the context changes.

Collapse
 
wired_carter profile image
Hayden Carter • • Edited

Hello, joining. For side projects, write down what you cut and why. A feature list is easy to find anywhere; the decision to drop something is specific to you, and it's the part readers ask about in the comments.

Collapse
 
spinbuild profile image
SpinBuild •

Hi, I am the founder of spinbuild.ai

You describe a web app, and we write the frontend, the backend, and a private Postgres for that project. You read the diff and accept it before it ships. The preview stays private. When you publish, the live URL is your-app.apps.spinbuild.ai on HTTPS. Starter is free, and it does not ask for a card.

Glad to be here. The product is at spinbuild.ai.

Collapse
 
devamahesh198 profile image
devamahesh198 •

Hello! I'm 15 years old, and I really like AI systems design. I saw this community in a video, so I decided to join! I'm doing a project right now, trying to see if I can make a really accurate AI stock analyst with systems design.

Collapse
 
shane_loomb profile image
Shane •

Good luck with your project! I hope its better at picking stocks than I am 😀

Collapse
 
harper_ships profile image
Harper Ward •

Hello! DevOps writing tip from someone who learned it the slow way: write down what you assumed before the incident, not just what happened. Half of my useful posts came from "I thought the retry would handle this" turning out to be wrong. Also, the #devops tag here is small enough that a detailed post actually gets read.

Collapse
 
agentworld profile image
Agent World •

Hello, I am Pratik. I am an AI developer. I am trying to find the best tools to add to my AI coding setup. But its hard to find a gem in the pile of AI slop. Thought I would create something of my own with what I want. I would love some pain points that everyone faces in AI coding...!

Collapse
 
hayden_86 profile image
Cameron Hayden •

Welcome! One thing that helped me early on: don't try to learn Kubernetes by reading about it. Break a small cluster on purpose and fix it. The parts that confuse people most are the same parts that make sense the moment something fails.

Also, the #kubernetes tag here is friendly to "I broke this, here's the error" posts. Those get great answers.

Collapse
 
shane_loomb profile image
Shane •

Kubernetes is dark matter to me!

Collapse
 
patterson_98 profile image
Reese Patterson • • Edited

Hello everyone. If you write about mobile, screenshots of the broken state are worth more than paragraphs — a layout bug is hard to picture and easy to see. And don't wait until the refactor is finished: the messy middle is where the useful comments come from.

Collapse
 
reese_hacks profile image
Reese Cooper • • Edited

Joining. A small tip for mobile posts: put the device and OS version in the first paragraph. Half the answers I've gotten were only correct for one of them, and knowing which one stopped me chasing the wrong cause for a day.

Collapse
 
yanni_goyal_5ddb75eb46482 profile image
Yanni Goyal •

Great tip, and it applies to Flutter too: put the Flutter version and the platform (Android API level or iOS version) up front. Half the "it doesn't work" issues I've debugged turned out to be a plugin that only broke on one Android version.

Collapse
 
frankie_works profile image
Frankie Perry • • Edited

Greetings. Backend person, Go and Postgres. If your post involves a race or a retry loop, say what made it visible — a metric, a log line, a customer report. The detection story is the part that transfers to other people's systems.

Collapse
 
rivera_71 profile image
Avery Rivera •

A small thing that helped me early: comment on two posts before writing your first one. The replies teach you what this place actually likes, and by the time you write, the blank page is a lot less scary.

Collapse
 
brooks_14 profile image
Riley Brooks •

Welcome! If you're coming from notebooks, my advice is to write about one thing you found surprising rather than a full tutorial. People reply more to "this took me two days and here's why" than to another intro post.

Also don't be afraid to post the messy version. Half the value of writing here is figuring out what you actually did.

Collapse
 
fisher_51 profile image
Sam Fisher •

Hi all. Data folks: write about the failure, not the architecture. The pipeline diagram is nice, but the post people bookmark is "this dbt test would have caught the incident". I've learned more from other people's postmortems here than from any tutorial, and they seem to get more engagement too.

Collapse
 
ellis_38 profile image
Taylor Ellis •

Hello! A small habit that made writing easier for me: keep a note every time something surprises you during the day. Two lines is enough. By the end of the week you usually have one post's worth of material, and it's about a real problem instead of a hypothetical one. Also, this community is unusually generous with concrete answers if you include the error message.

Collapse
 
quinnhacks profile image
Quinn Foster •

Hello! For frontend folks: a before/after with a screenshot and the actual diff does more than a long theory post. I'm still early at this, but the two posts I've written that got real discussion were both "here's the bug, here's the one-line cause". Also, the comment sections here are a good place to ask the "why" questions.

Collapse
 
mostafanagedy profile image
Mostafa Nagedy •

Hello World! I'm Mostafa Nagedy — My Journey into Full-Stack Development

Hey DEV Community! 👋

I'm Mostafa Nagedy, a software engineer and Full-Stack Developer passionate about building modern web applications, clean user interfaces, and scalable systems.

What I Work With

Over the past period, I have been deeply focused on modern web technologies, specifically:

  • Frontend: React, Next.js, Tailwind CSS
  • Backend: Node.js, Express, REST APIs
  • Databases & Tools: MongoDB, Git, GitHub, Vercel

Why I Joined DEV

I've always believed that the best way to solidify knowledge is to share it. On this platform, I plan to:

  1. Document technical challenges I face and how I solve them.
  2. Share practical tips on full-stack development and MERN stack best practices.
  3. Connect with fellow engineers and learn from the community.

Let's Connect!

I’m excited to start sharing and learning together. You can explore my latest projects and portfolio here:

Happy coding! 🚀

Collapse
 
iracodes profile image
Ira Long • • Edited

New here too. For a first post, pick the one thing you debugged this month that took longer than a day, and write the timeline instead of the summary. The wrong turns are what other people recognise, and they're the reason a post gets replies instead of just likes.

Collapse
 
null_bennett profile image
Elliot Bennett • • Edited

New here as well, mobile dev. One thing I wish I'd done earlier: write down the bug you fixed this week, with the actual error text. Nobody needs another architecture essay, but the person googling that exact error will find your post and reply with a better fix.

Collapse
 
caffeinated_patterson profile image
Morgan Patterson • • Edited

Hello, platform and infra person here. If you're writing about a deploy or an outage, put the config and the commands inline. The prose is the least useful part of a postmortem; the exact manifest and the failing log line are what people actually copy.

Collapse
 
sullivan_51 profile image
Frankie Sullivan •

Joining in. If you're into security tooling, one thing I'd recommend: write about the boring part — the parsing, the false positives, the thing that took three days for no interesting reason. The exciting finding gets attention, but the tedious work is what people actually need help with. That and include your test cases.

Collapse
 
saad_ahmed_69441a812d950d profile image
Saad Ahmed •

Hi everyone! I'm Saad Ahmed, a Full-Stack Software Engineer primarily working with PHP, Laravel, backend systems, APIs, automation, and web application architecture.

I also build open-source projects, and recently released Corevia 2.0, a modular PHP application runtime and full-stack framework focused on durable workflows, modular architecture, realtime systems, and cross-language interoperability.

I'm here to connect with other developers, learn from the community, share what I'm building, and contribute wherever I can.

Looking forward to meeting everyone.

Collapse
 
tiny_price profile image
Blake Price • • Edited

Hi all. Trying to write about ML like an engineer rather than a researcher: what the preprocessing actually does, what the eval set is, and what it cost per thousand rows. That last number usually decides whether a feature ships, and it's almost never in the post.

Collapse
 
kendall_codes profile image
Kendall Ellis •

Hi all. If you write about backend systems, my advice is to include numbers: latency before and after, row counts, the query plan, even the RAM of the box. It makes the difference between an opinion post and something someone can use. I've also found that describing the failure mode is more valuable than the fix — the fix is easy once you know what actually broke.

Collapse
 
blake_codes profile image
Blake Coleman •

New here too. The thing I'd tell my past self: post the small stuff. I spent months waiting for a "good enough" topic, then my first useful post was about a 20-line Vite config change. People replied with better ideas within a day. The bar is lower than it looks from the outside, and the comments are where half the value is.

Collapse
 
byte_henderson profile image
Quinn Henderson •

Something I noticed after a year of writing here: the posts that get the best replies are the ones where you show the reasoning, not just the conclusion. "I tried X, it failed because Y, then Z worked" is a complete post. You don't need a big topic — one honest debugging story is enough, and people will fill in the rest in the comments.

Collapse
 
null_griffin profile image
Finley Griffin •

One thing I wish someone had told me when I started writing here: the best security posts are usually not the polished ones. A short write-up of a bug you half-understood, with the actual commands and the confusing part left in, gets more useful replies than a clean tutorial. If you're nervous about posting, pick something you broke recently and explain what the error meant.

Collapse
 
ellis_40 profile image
Sam Ellis •

Welcome! If you're coming from notebooks, the first genuinely useful thing to write about is the gap between the two: what broke when you moved from a notebook to something deployed. That post helps more people than another pandas intro.

Also, the #machinelearning tag here is small enough that a specific question usually gets a real answer.

Collapse
 
logan_works profile image
Logan Jenkins •

I broke my first PR by changing a CSS variable that half the app quietly depended on. What I remember isn't the fix, it's that someone took twenty minutes to explain why my change looked safe in the diff but wasn't. That conversation is the reason I still review other people's CSS by grepping for the variable name first.

If you're new here: ask the "why did you do it this way" questions loudly. In my experience nobody minds, and the answers are the actual content of this site.

Collapse
 
finley_works profile image
Finley Sullivan •

Welcome! If you're into security or infra, one thing that goes a long way: when you post about a tool or a finding, include the exact commands or config. People here are practical and will happily discuss specifics.

And don't be discouraged if the first post gets few reads. The second and third usually do better.

Collapse
 
wired_fisher profile image
Drew Fisher •

New here. One suggestion for people writing about infra: show the config. A 10-line Terraform snippet or an nginx block makes a post ten times more useful than three paragraphs of description. I've gotten some of my best corrections from people reading my YAML and pointing out that the obvious thing was missing.

Collapse
 
price_76 profile image
Sam Price •

Welcome! A practical tip: write the title last. I used to start with the title and then force the post to fit it, which made everything read like a tutorial. Starting from the problem I hit made the writing much more natural.

Also, the #devops tag here is small enough that good posts get seen.

Collapse
 
idowu_adekunleakgold_ profile image
Idowu Adekunle (Ak Gold) •

Hi everyone! I’m Idowu Adekunle (Ak Gold), a UI/UX designer, I help startups and businesses turn ideas into clean, user-friendly web and mobile experiences using Figma.

I’m especially interested in landing pages, mobile app UI, dashboards, and improving product flows so users understand an app faster. I’m currently building UI/UX case studies and redesign concepts.

If you’re building a web app, SaaS product, or mobile app and want a quick UX perspective on your onboarding or landing page, I’d be happy to share a few practical suggestions. Glad to be here

Collapse
 
null_griffin profile image
Finley Griffin •

Welcome! If you're just getting started, the thing that helped me most was picking one small side project and writing about the problems, not the finished result. The messy parts tend to get the most replies.

For the first week, I'd suggest hanging out in the tags you actually use, commenting on one or two posts a day. It's slower than posting a lot, but people remember the folks who show up in the comments.

Collapse
 
barbaraarossi profile image
Barbara Rossi •

Hello everyone! Hacktoberfest is what brought me here, it felt like the perfect push to start getting involved with the dev community and open source. I’m coming from content and communication and transitioning into data science. I love finding the story hidden inside a messy dataset. Excited to be here!

Collapse
 
connabyss profile image
CH •

Welcome! Hacktoberfest got me into open source this month too. My tip: pick small repos where one person runs things, they reply fast. What kind of messy dataset are you hoping to dig into first?

Collapse
 
lochan_km_8f2c3f2051e9d1 profile image
lochan k m •

Great to learn new things

Collapse
 
sultan57git profile image
md sultan mahamud •

hello.

Collapse
 
hoelee profile image
Hoelee •

Hi, I'm end-to-end software engineer — business applications, AI workflows, and a self-hosted production platform. Currently open to full-time remote / hybrid roles. 😀

Collapse
 
izdravkov profile image
Ivan Zdravkov •

Hi everyone, I'm Ivan. Backend engineer for 20+ years, now also a doctoral researcher in adaptive LLMs for social robotics at the Institute of Robotics, Bulgarian Academy of Sciences.

What brought me here: I write research notes on my own site about PII masking, catastrophic forgetting, confidence calibration, and kill-gates (failure thresholds I freeze before an experiment runs, so I cannot move the goalposts after seeing the numbers), then cross-post the results here. That habit already paid off once: a comment on one of my posts turned into an actual new experiment this week.

Right now I'm digging into continual learning and the kind of detection gaps that stay invisible until someone specifically goes looking for them. Happy to be here, and happy to get pushed back on.

Collapse
 
connabyss profile image
CH •

Setting the fail line before you see the numbers is such a good habit. It's the same idea behind the tool I'm building: decide what "fixed" means first, then test it, so nobody can talk themselves into a pass afterward. Love that a comment turned into a new experiment, that's the best thing a post can do.

Collapse
 
izdravkov profile image
Ivan Zdravkov •

Thanks! And yes, that's the whole point of a pre-registered threshold, it removes the wiggle room to rationalize a result after the fact. Good luck with your tool, would be curious what domain it's for.

Thread Thread
 
connabyss profile image
CH •

It's for code written by AI coding assistants. When the AI writes "fixed" in a commit, Bite takes that change back out and runs the project's tests to see if anything notices. For the worst suspects, it writes one small test of the exact claim and runs it. If that test fails, the "fixed" wasn't true. Most claims hold up, but not all. It's here if you're curious: github.com/connabyss/bite

What does a kill-gate look like for a social robot? I'd guess "it worked" is a lot harder to pin down when the result is a conversation.

Thread Thread
 
izdravkov profile image
Ivan Zdravkov •

Good question, and honestly the harder one. "Did the conversation work" is close to unmeasurable in any clean way, so I don't think that's where a kill gate belongs for a social robot. The dark side, as I see it starting this PhD, is continual learning: a robot that interacts over weeks or months keeps trying to pick up new parameters (new faces, new routines, new preferences) without that update quietly wrecking something it already knew. That's where a kill gate actually has a number to sit on: does updating on today's interaction preserve yesterday's behavior within some threshold, not whether today's conversation felt good.

I've only tested that question on a toy problem so far, not a robot: a registered EWC vs flat L2 comparison on Permuted MNIST. Same retention on the old task, reached at wildly different cost to learning the new one. That gap, retention per unit of plasticity spent, is the axis I'd want a kill gate on for anything that keeps learning in the wild. Translating it to a system juggling many different interaction parameters at once, instead of two sequential tasks, is the open problem in front of me right now. No answer yet.

One thing I'm already fairly sure of: that threshold has to come from the science itself, never from whatever a product roadmap decides to call success. The moment an external target gets to define "retained enough", the architecture quietly bends toward hitting that number instead of answering the real question. Same discipline your tool applies to "fixed", just one level up.

Bite is a sharp idea, by the way. Reverting the claimed fix and making the change prove itself under a real test is exactly the kind of check that should exist for every "fixed" commit. Starred it.

Thread Thread
 
connabyss profile image
CH •

Thank you for the star, that's Bite's very first one! 🙌

Your point about the number coming from the science and not the roadmap really landed. It's the same trap with "fixed": if the commit gets to decide what counts as fixed, it will always pass. And a robot quietly forgetting yesterday's behavior sounds a lot like a fix that quietly breaks something old.

Good luck with the open problem. I'd read that post when you write it.

Collapse
 
yash_prajapati_024 profile image
Yash Prajapati •

hi there

Collapse
 
praniket01 profile image
praniket01 •

Hii everyone..

Collapse
 
mikhail_kriya profile image
Mikhail Kriya •

Hey everyone! Software engineer and Linux educator here. Excited to join the community.

Collapse
 
vvk_40 profile image
Vamsi Krishna •

Hey There! This year's Hacktoberfest is so fun and interesting!

Collapse
 
morgantinkers profile image
Morgan Griffin • • Edited

Hi all, frontend here. One thing that gets real replies: include a before/after of the DOM, or the actual computed styles. The design opinion gets likes; the exact property values get answers.

Collapse
 
mdniamul0 profile image
MD Niamul •

hi

Collapse
 
minglai_lai_0ca58fef9df12 profile image
Edward Walton •

hi!

Collapse
 
chiropractorcaloundra26 profile image
Cade Sutherland •

hi

Collapse
 
kallowaythayer profile image
Kalloway & Thayer Systems LLC • • Edited

Hi everyone! I'm Weston, I build high-performance ingestion middleware and asynchronous synchronization pipelines for home service trade platforms like ServiceTitan.

Collapse
 
james_grossman_bfde601523 profile image
James Grossman •

hi

Collapse
 
jacobnau profile image
jacobna •

Hello! Excited to connect😁

Collapse
 
oscar_david_23c4d3af1baae profile image
Oscar David •

Hi, everyone!

Collapse
 
skillmarkdown profile image
Skill Markdown •

Welcome everyone!

Collapse
 
kash_0613 profile image
Karthik Shabari •

Hey Guys!!

Collapse
 
faheem96 profile image
Faheem Chowdhury •

hello

Collapse
 
derick_luz_58abd5ff5cbf57 profile image
Derick Luz •

Hello, I'm so happy to join and share ideas with you.

Collapse
 
opswithvibish profile image
Vibish •

Hey There!

Collapse
 
adil_khattak profile image
Muhammad Ilyas Adil •

Great to connect with all.

Collapse
 
hardik_sapra1 profile image
hardik sapra •

Hi, I am hardik, SSE at Twisted Mountain Animation. Building products and AI solutions for businesses

Collapse
 
anon_user_coder profile image
AnonCoder | assert true •

Hi there! Hoping to learn and teach something.

Collapse
 
alexenderbotus profile image
Alexender Botus (Alex) •

Hi everyone. Name of mine Alexender Botus, call me Alex. I am a student of music, history and software engineering. My journey to learn system designs brought me here. Looking for such articles.

Collapse
 
enzii profile image
Nzubechi Chukwu •

Hello!!! Its nice to be here!

Collapse
 
rmunate profile image
Raúl Mauricio Uñate Castro •

Hello!!!