Skip to content
360SOFTY
HomeInsights

Insights

Ideas for building
better software.

Engineering perspectives on architecture, product development, AI, and the everyday decisions behind useful software.

From the engineering desk

Browse practical articles or follow the latest technology updates.

Subscribe via RSS →

Curated links from external sources — not 360Softy original articles.

ExternalSoftware Engineering
DEV Community

Optimizing RAG at Scale: Chunking, Retrieval, and the Bayesian Search That Cut Latency 40%

Optimizing RAG at Scale: Chunking, Retrieval, and the Bayesian Search That Cut Latency 40% How we moved from "semantic search + hope" to a measured, tunable retrieval pipeline with 95% recall@10 Everyone ships RAG the same way: chunk by 512 tokens, embed with text-embedding-3-small, top-k=5, stuff into context. It works for demos. Then you hit production: Legal contracts: 512 tokens splits clauses mid-sentence API docs: 1000-token chunks drown signal in noise Customer tickets: Conversational c

llmaievaluation
DEV CommunityRead original
ExternalSoftware Engineering
DEV Community

A Dart regex that cannot be made to hang

FFI binding to Google's RE2. Matching runs in linear time, so no catastrophic backtracking. Here is a validator that ships in a lot of Dart code. It checks that a string is a list of words separated by single spaces: final ok = RegExp(r'^(\w+\s?)*$'); It works. Type a name, a tag list, a short label, and it returns true or false quickly. Then someone sends 31 characters of x with no trailing separator and the isolate stops answering. I timed it. On Apple Silicon, Dart 3.11, dart:core RegExp tak

dartflutterregex
DEV CommunityRead original
ExternalSoftware Engineering
DEV Community

How Does an LLM Request and Response Cycle Work? A Full Walkthrough

In this blog post, we will see how an LLM request and response cycle works, from the second you hit send to the moment the last word lands on your screen. I will not throw a wall of transformer math at you. Instead, we will follow one single prompt through every stage of the journey, so nothing feels abstract. A single prompt sent to a large language model passes through multiple distinct stages before a response appears on screen. The process begins with the client app building a structured JSO

aibeginnersllm
DEV CommunityRead original
ExternalSoftware Engineering
DEV Community

When a human edits an agent's output, where does that decision go?

An agent drafts a reply to an insurance claim. A human reads it, changes one sentence that would have committed the insurer to something it did not mean, and ships it. That edit is the single most valuable thing that happened in the whole interaction. It is a human catching an error a model could not. Now ask where that edit is recorded. In most production systems, the answer is: in application code, a chat thread, a ticket comment, or nobody's memory at all. The judgement that mattered most lea

aiopensourceshowdev
DEV CommunityRead original
ExternalSoftware Engineering
DEV Community

I Tried Kimi K3 for a Week. Here’s What Happened.

A few days ago I posted my first impressions of Kimi K3 on LinkedIn. Since then, I’ve been using it daily for real work — not benchmarks, not demos. Here’s what actually held up. The Context Window Isn’t Marketing Frontend Coding: The Hype Checks Out It’s also #1 on nextjs.org/evals — the first open model to beat all proprietary ones on that benchmark. Agentic Work: Good, Not Magic The Honest Truth Why the Open Weights Matter Bottom Line If you want to try it: kimi.com P.S. — Original first impr

aiprogrammingdevelopment
DEV CommunityRead original
ExternalSoftware Engineering
DEV Community

We built Multi-Repo Workspaces for AI coding

Most production applications aren't monorepos. A feature often spans the frontend, backend, shared packages, infrastructure, and more. While polyrepos work well for ownership, AI coding tools typically only see one repository at a time. We wanted to change that. With Multi-Repo Workspaces in DevSwarm, you can bundle multiple Git repositories into a single workspace while keeping each repository independent with its own history, branches, remotes, CI, and pull requests. Your AI agent, whether it'

devtoolsvibecodingcoding
DEV CommunityRead original
ExternalSoftware Engineering
DEV Community

Optimizing RAG at Scale: Chunking, Retrieval, and the Bayesian Search That Cut Latency 40%

Optimizing RAG at Scale: Chunking, Retrieval, and the Bayesian Search That Cut Latency 40% How we moved from "semantic search + hope" to a measured, tunable retrieval pipeline with 95% recall@10 Everyone ships RAG the same way: chunk by 512 tokens, embed with text-embedding-3-small, top-k=5, stuff into context. It works for demos. Then you hit production: Legal contracts: 512 tokens splits clauses mid-sentence API docs: 1000-token chunks drown signal in noise Customer tickets: Conversational c

llmaievaluation
DEV CommunityRead original
ExternalSoftware Engineering
DEV Community

Building Production-Grade LLM Evaluation Pipelines: From Vibes to Metrics

Building Production-Grade LLM Evaluation Pipelines: From Vibes to Metrics How we replaced "looks good to me" with automated evaluation catching 92% of hallucinations before deployment Three months ago, our team shipped a RAG-based customer support assistant. It worked great in testing — we'd ask it questions, read the answers, and say "yeah, that looks right." Then it hit production. A customer asked about their billing cycle. The assistant confidently cited a policy that didn't exist. Another

llmaievaluation
DEV CommunityRead original

Let’s start with a conversation

Tell us what you’re working on.

An idea, a challenge, or a system that needs to work better. We’ll help you understand the next step.

Discuss your project

Prefer email? [email protected]