Pinned
Alvin Sng
949 posts
Alvin Sng
@alvinsng
Our launch caused record-breaking traffic. TBH, I'm a bit disappointed. I was expecting sleepless nights, firefighting multiple site outages, and a heroic bug fix for a legendary story in the office. Instead, our #serverless stack just... worked. Missed out on the classic- Our newest engineer shipped 3 PRs on their first day. That's our normal. @FactoryAl takes dev productivity seriously. How we achieve it: - Dead-Simple Setup: git checkout, nom install, then npm run dev - Agentic Coding: Zero tribal knowledge needed to touch the code - Use SSO
- Don't be fooled by companies selling fancy SaaS products for AI code reviews. It does not need to be complicated. At @FactoryAI, our approach is to first build the top-ranking coding agent as the foundation (#1 on Terminal Bench). With that in place, a code review bot is just aWe set up a GitHub Workflow for PR reviews using droid exec (@FactoryAI’s headless CLI). It's already blocked real bugs from merging into prod. Workflow code in thread below.
- A lint rule a day keeps the agents from going astray. At @FactoryAI, we're continuously generating custom lint rules whenever we think of them. Don't waste time wrangling your agents when you can define the rules ahead of time. Read more about this on our blog:
- Here’s an inside peek at the assembly line of @FactoryAI, where we’re building the droids All captured with Meta glasses for a true POV experience. #DayInTheLife
00:00 - "Slop" unit tests are better than no unit tests: The @FactoryAI droid recently caught a bug that was undetectable from viewing the source code but was detected by looking at its own unit tests. Unit tests help agents catch harder bugs by validating runtime outputs. In this
- Awkward moment of the day: "Have you been here before?" - reception "Yeah, I've worked here before" - me
- Devs spoke, we heard. Hybrid model execution now availableThe Factory CLI now supports mixed-model runs. You can build a specification with a SOTA model, and execute with a more cost-efficient model. Developers deserve choice. You can now get the best out of all models, from every provider. For example, Sonnet 4.5 for specification
- Replying to @siranwrapperHow do you compliment a software engineer? Say you don't look like a software engineer
- This powers the mass refactors I do at @FactoryAI I spawn parallel droids to create PRs that touch hundreds of files in under an hour. Recently, I've: - Enforced that routes wrap in proper middleware - Fixed error handling to either log or throw, but never both - Conformed
00:22Starting today the #1 software development agent in the world is fully scriptable. For the past few weeks I've been working on droid exec, Factory's headless CLI for agent native automation. Droids can now work independently, without human input. Create new autonomous workflows - Love it when I can drop a link to failing CI checks, and it will just "go figure". Building @FactoryAI with Factory 🔁
- AI coding tip: Use LLMs to get advice on naming. They say one of the two hardest problems in computer science is naming. However, LLMs actually excel at this since they "weight" the most common conventions as their bread and butter. The next time you're thinking about how to name
- Starting today, you can use any open-source model to power your Droids. Droids achieve the highest scores across all open-source models on Terminal-Bench. We find GLM 4.6 to be the most performant, remarkably achieving a score in Droid that beats Sonnet 4 in Claude Code.
