Qwen 3.8 27B is excellent, but it defaults to overthinking things
Article URL: https://simonwillison.net/2026/Aug/16/qwen-38-27b/ Comments URL: https://news.ycombinator.com/item?id=49324985 Points: 751 # Comments: 355
Insights
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.
Article URL: https://simonwillison.net/2026/Aug/16/qwen-38-27b/ Comments URL: https://news.ycombinator.com/item?id=49324985 Points: 751 # Comments: 355
Every Hydrogen project I've worked on started the same way — wire up a cart, build a PDP, add filters, then spend weeks on the "extras" that aren't really extras: wishlist, compare, quick view, proper i18n, RTL. After doing this enough times I decided to build it once and properly. ada ÉLAN is a Hydrogen storefront theme for fashion brands. It ships: — An editorial design system (Cormorant Garamond + Plus Jakarta Sans, documented design tokens) Stack: Hydrogen 2025.7, React Router 7, Tailwind 4,
Nintendo had a hotline where gamers could, at the time, call and speak with 'Game Counsellors' who provided them with tips and walkthroughs. It operated for quite sometime before Nintendo sunset it. There are a few (Product) lessons from this that I am sure will be of value to Product Leaders. 1- Necessity (Invention's mother): The necessity of a situation usually births the creation of something that stands out from the rest. While Nintendo was not the first to use a phone as a 'business' funct
This is a submission for the DEV Weekend Challenge: Dog Days Edition. Same sidewalk. Two bodies. Two completely different heat loads. Banx is my French Bulldog. Born October 5, 2022. He weighs 35 pounds — seven above the 28-pound ceiling in the French Bull Dog Club of America conformation standard. I call him my XL. He is purebred and he has never had airway surgery. The face that makes him Banx is also the conformation that puts French Bulldogs at higher risk of obstructed breathing and heat-r
AI Prompt Lab is a full-stack application I built around a simple idea: managing reusable AI prompts should feel like working with any other structured application asset, rather than keeping them scattered across notes, text files or chat histories. The project combines Java 21, Spring Boot, React 19, TypeScript, PostgreSQL and OpenRouter in a compact architecture that covers the main concerns of a modern web application: authentication, authorization, persistence, external API integration and m
Most developers think of a PostgreSQL table like this: 1 | 25 Then we run: UPDATE users The obvious mental model is: 25 → 30 But PostgreSQL doesn't simply overwrite the old value. To understand what actually happens, we need to look at how PostgreSQL stores rows and how MVCC (Multi-Version Concurrency Control) works. Tables are stored as pages PostgreSQL stores table data in fixed-size pages. The usual page size is 8 KB. You can roughly imagine a table as: users ┌─────────────────┐ Each page ca
AI is changing accessibility remediation fast. It can find common accessibility failures, explain them, and increasingly write the fix. That's real, and it's useful. But finding a failure is not the same as deciding what the fix should be. The gap is easy to miss because it doesn't always produce an obviously wrong answer. Sometimes it produces a confident, plausible fix that is wrong for reasons that only become apparent when you understand the interaction in context. I recently worked through
How VIDRAFT Hit 510.58 TPS on Gemma-4: A Deep Dive into "The First Gemma Challenge" Win TL;DR: VIDRAFT topped "The First Gemma Challenge" leaderboard with a verified 510.58 tokens-per-second (TPS) score on google/gemma-4-E4B-it using a single NVIDIA A10G GPU — while a rival submission posted a faster raw number but failed the quality gate. This post breaks down the public configuration choices that made it possible, and what engineers can steal for their own inference tuning work. "The First
Let’s start with a conversation
An idea, a challenge, or a system that needs to work better. We’ll help you understand the next step.
Prefer email? [email protected]