Atheros

I always believed in innovative ways for learning, here is a vizualization of how CSS grid work 🚀

VEED just launched a dubbing API with lip sync, and this one is personal for us. A huge part of our Atheros Learning community speaks Spanish, but we never found a good enough dubbing solution. Every dub looked off, and we were never satisfied with the results. Super excited to build with VEED’s API and finally bring our courses to learners in multiple languages.

Decision models are one of the most exciting areas in AI this fall, but without retrieval they can only judge what’s already in their context. Web grounding was the missing piece, and the Parallel Decisions API closes that gap. I can already see where this fits in Atheros products: gates in our agent loops that check the latest changelogs for deprecated APIs, grounded judges that flag stale content in our courses, and our frontend agent knowing whether a library version is still current. Checks that were too slow and expensive with full LLM calls now cost a fraction of a cent per decision. Can’t wait to start using Parallel in our stack! #ParallelPartner

Agentic coding vocabulary for iOS navigation 📱 Be sure to know the name to build more precisely 🧑‍💻

You can learn AI and machine learning completely for free 🤖 Here are my top 3 resources 🧑‍💻

In this agentic coding vocabulary, we are going over web page layouts fundamentals🧑‍💻

Please welcome the newest member of the Atheros team. Viktor joined as a frontend engineer, and in his first month he helped our designers ship 17% more UI fixes. He’s also an AI employee. Viktor lives in our Slack with integrations to 9,800+ tools, including GitHub, and our designers started handing him UI fixes the way they would to any engineer. They ship without writing code, and he sends screenshots so they can check each change. For years I treated AI like any other software, so I was skeptical at first. With Viktor we did what we’d do with any new hire. We gave him feedback, he learned from his mistakes, and we started delegating more complex tasks, with our leadership’s approval. One of the main flaws of agentic frontend coding is that agents tend to duplicate UI components that already exist. There are ways to reduce this, like repo maps, but Viktor proactively catches PRs that do it, which keeps our React frontend modular and in line with our design system. Our designers keep shipping, his first-pass PR reviews save our engineers tens of hours a week, and our PMs check decisions with him against GA traffic, IG analytics and our SES email stats in AWS RDS. He started as our designers’ teammate, and now he works with everyone. Viktor isn’t a tool we use. He’s someone we work with. If you want @viktor_com on your team, you can start for free. Link in the first reply. #AIemployee #AIengineering Paid Partnership

In this agentic coding vocabulary, we are going over iOS gestures fundamentals 🧑‍💻 This is just part 1. of the series, many more to come 📱🚀

Astra sidebar built with HTML, CSS and JavaScript, but also with ReactJS 🧑‍💻 Code 👇

AI coding agents offer incredible performance boost, but there is a big code quality trade off 🧑‍💻 This is the first part of series, where I show common mistakes that AI agents keep making 🤖

Loading patterns for building better products 🚀 If you know the pattern, you can prompt more precisely 🧑‍💻

When coding with AI … it is very useful to know exact names of more complex animations 🧑‍💻

Our designers shipped 17% more UI fixes a day in the first month. None of them learned to code. And no engineer had to build them. I’ve always believed AI belongs right next to where the team talks, executing the moment a decision is made and a task approved. We build AI driven products, and every small change in UI we need to ship means a ticket and engineering work. But engineers should focus on complex challenges, not simple UI fixes. We tried Claude Code, but it lives outside of the main communication channels, so designers still handed every fix to an engineer. Viktor closed that gap. He joined our Slack workspace as one of our teammates. A designer describes the change, he opens the PR on GitHub right away, and an engineer reviews and merges. In the first week I nearly switched him off, but by week 2, our engineers had tens of hours a week back for client work, helping us ship higher quality, faster. Then the rest of the team moved in. With 3,200+ integrations ... PMs, engineers and designers are now using Viktor to validate ideas on the spot against real analytics as well as delegate tasks to Viktor. Teams don’t need another AI to switch to. They need one already integrated in their workflow. Most AI tells you what to do. Viktor does it. Hire @viktor_com for your team. $100 in credits included, no card. Full link in first comment. #AIemployee #AIengineering Paid Partnership

Magazine layout built with HTML and CSS🧑‍💻 Code 👇

MacOS Dock inspired by liquid glass 🖥️

Subquadratic just launched SubQ Snap, a decision model with 384K multimodal context. That's 12x Jev's context, and it beats Jev on the Decision Index too. Decision model as a judge is one of the best use cases I see for us. Until now, Jev's 32K text-only context could be limiting for long or multimodal inputs. Super excited about SubQ Snap 🚀

My top 15 VSCode extensions for AI workflows in 2026 🚀

Here is the overview of different scrolling patterns 🧑‍💻🚀 Often useful for AI assisted coding to precisely describe what you need 🤖