Tom MacWright
@macwright.com
6 months ago there are some date representations with epochs near 1850, like mumps and vms, but not cobol en.wikipedia.org/wiki/Epoch_(...
💬 1
♻️ 0
❤️ 0
there are some date representations with epochs near 1850, like mumps and vms, but not cobol en.wikipedia.org/wiki/Epoch_(...
docs from ibm about 85 cobol date storage: www.ibm.com/docs/en/cobo...
sorry but like, this isn't right? iso 8601 is a spec for text representations of datetime. it doesn't have an epoch and certainly not one that is 1850. 85 cobol does have dates stored from an epoch, but it's either 1601 or 1582
We released a tool that converts Xcode projects into XcodeGraph, enabling features like selective testing and graph visualization for all users: tuist.dev/blog/2025/0...
Tuist's server pricing is straightforward:
- Light use? It's on the house.
- Need more? Pay only for the value you get.
Your negotiation skills will never be taken for granted.
Discover more: tuist.dev/pricing
For more details on edit prediction and Zeta, read our full blog post:
zed.dev/blog/edit-pr...
As with the rest of Zed, we're embracing an open approach to the development of AI features. Help shape the future of dev tools: contribute to Zeta's dataset, explore its code, or deploy it yourself.
huggingface.co/datasets/zed...
For further reduce latency, we've deployed GPUs in North America and Europe, with more regions coming soon. Cloudflare workers route your requests to the nearest data center, ensuring a responsive experience worldwide.
To meet our strict latency targets, we leveraged speculative decoding. This technique parallelizes token generation by using the input as a reference, significantly speeding up predictions without sacrificing quality.
We fine-tuned Zeta using Direct Preference Optimization (DPO), teaching it both good and bad edit patterns. This allows us to deliver high-quality predictions that avoid common pitfalls, even in tricky edge cases.
Our evals for Zeta use a larger LLM to assess if Zeta's suggested edits match our intent. This allows us to validate predictions even when outputs vary, ensuring consistent quality across different coding styles.
Zeta predicts edits at arbitrary locations in your code. It rewrites larger chunks around the cursor, enabling natural, context-aware predictions that go beyond simple completions.
Edit prediction is powered by Zeta, our new open-source model derived from Qwen2.5-Coder-7B and fine-tuned with an open dataset. We hope you'll contribute data while editing open source repositories in Zed to help improve it further.
huggingface.co/zed-industri...
Edit prediction also meshes cleanly with completions from your language server. When the LSP-based autocomplete menu is visible, you can hold `option` or `alt` to preview the predicted edit, hiding the menu for an unobstructed view.
Try edit prediction for free during our public beta. Download Zed and sign in with your GitHub account to experience it firsthand. We look forward to your feedback!
zed.dev/edit-predict...
As you work, Zed predicts your next edit. Apply it with a simple 'tab' press, or chain multiple edits by pressing 'tab' repeatedly. It's a natural extension of the Zed experience, saving you time and keystrokes.
We engineered Zed to be instantly responsive, but what's faster than instant? Anticipating your next move. Introducing edit prediction, powered by Zeta, our new open-source language model. 🚀
Discover how Ruby and Ruby on Rails prioritize joyful design in our latest newsletter. Learn why designing for joy matters in app development: tuist.dev/newsletter/...
Subscribe: tuist.dev/newsletter/...
I’m in love with work. In parallel we are also working on making all of this easy to integrate into your Xcode projects (just one command)
There's a ton going on in the world of AI coding agents 🤖
We're building the future of serverless databases with Prisma Postgres and it's the perfect fit for these agents 🤝
ICYMI, check out how we are solving the database layer for AI coding 👇
pris.ly/ai-agents
We're crafting a top-tier productivity platform tailored for teams using Xcode, with plans to expand to Android, Flutter, and React Native. As part of this mission, we're transforming our dashboard to align perfectly with our vision. What do you think about our upcoming design?