@LucasGoldner 1- Liquid Glass support is already but without an official support, anyway they will support it gradually after the big decoupling of the material/cupertino widgets
2- macros not so important as AI will write models anyway and refactoring fast
3- Fast Flutter Web YES YES YES
@zachlloydtweets As a mobile developer, Iโd love to work with your team to build a software factory for solo developers. DM me if youโd like a mobile developerโs perspective
Without AI, this growing complexity could make software increasingly expensive, forcing customers to either pay more or settle for lower-quality products.
Maybe AI isn't just replacing engineering work. Maybe it's making modern software engineering economically sustainable.
AI might have saved software engineering from its own complexity. ๐ค
Building even a simple website today can involve DevOps, SRE, Docker, Terraform, CI/CD, monitoring, and security.
Add AI features, and suddenly you're dealing with LangChain, Langfuse, MLOps, vector databases, and more.
Software engineering has become so complex that building and maintaining products can require entire teams of specialists.
The Full SDD Lifecycle (Github spec kit)
You can automate parts of the process using workflows.
You can also automatically sync everything with Jira if you want. However, I prefer to keep all documentation directly in the project repository instead of using Jira issues.
โก A tiny 1.5B model outperformed GPT-4.1 on the same benchmark.
The takeaway: Smaller, domain-specialized models can outperform general-purpose AI on targeted tasks.
Imagine what this could mean for Flutter, Laravel, and other specialized tech stacks! https://t.co/OUNWgy19fK
๐ Specialized AI can beat frontier models!
A new paper, Full-Stack Fine-Tuning for the Q Programming Language, reports impressive results:
๐ A specialized 32B model achieved 59% pass@1, outperforming Claude Opus 4 by 29.5% on the Q benchmark (Leetcode).