iiinova — where your digital self keeps evolving

Reposted by Nate Sesti

Ty Dunn @ty.energy · 17/02/2026

amplified.dev is a single-page essay arguing developers are amplified, not automated. One HTML file, inline CSS, vanilla JS. No frameworks. We wrote 11 checks for it. Checks don't have to protect code — they can protect taste. Here's what ours look like.

amplified.dev

Developers amplified, not automated

Every time a layer of the work became infrastructure, the role didn't shrink. AI is the next layer in it.

282

Reposted by Nate Sesti

Alan Nichol @alanmnichol.bsky.social · 04/09/2025

So 🔥 I love when startups do research and just ship in product. Skip the benchmarking theatre.

052

Nate Sesti @natesesti.bsky.social · 04/09/2025

This work was possible thanks to a generous compute grant from the Nvidia DGX Cloud Innovation Lab, allowing us access to a dedicated 8xH100 GPU instance for distributed training and inference

000

Nate Sesti @natesesti.bsky.social · 04/09/2025

Instinct was the product of an incredible summer internship project by Adarsh, with guidance from the rest of the Continue team We're just getting started with this first release and are particularly excited to push forward Next Edit by: - post-training with KTO - scaling the dataset by 10x

100

Nate Sesti @natesesti.bsky.social · 04/09/2025

Instinct's weights, dataset, and training code are all available for use under the Apache-2.0 license If you are interested in furthering the state of the art, either as part of the community or the Continue team, please reach out! huggingface.co/continuedev/...

huggingface.co

continuedev/instinct · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

120

Nate Sesti @natesesti.bsky.social · 04/09/2025

To evaluate Instinct, we adopted an LLM-as-judge strategy, similar to Zeta from Zed, asking Claude to rank edits on a scale from 1-5 based on functional equivalence to the ground truth edit Instinct’s average score of 3.877 outperforms Zeta’s score of 3.735

100

Nate Sesti @natesesti.bsky.social · 04/09/2025

Given that Qwen2.5-Coder-7B is already strong base model, we wanted to preserve its existing coding knowledge We used Selective Knowledge Transfer (SeleKT), which computes dense gradients, but applies weight updates for only the top-k gradients www.microsoft.com/en-us/resear...

microsoft.com

NextCoder: Robust Adaptation of Code LMs to Diverse Code Edits - Microsoft Research

Software engineering activities frequently involve edits to existing code. However, contemporary code language models (LMs) lack the ability to handle diverse types of code-edit requirements. In this ...

110

Nate Sesti @natesesti.bsky.social · 04/09/2025

Our dataset of 4,000+ high-quality real-world edits was built by collecting keystroke-level data from the Continue team and chunking it into coherent edits For multilingual support we used Qwen3-Coder-30B to "translate" these into synthetic data for Java, C++, Python, and Rust

110

Nate Sesti @natesesti.bsky.social · 04/09/2025

While traditional autocomplete inserts text by using a fill-in-the-middle prompt format (FIM), Next Edit accepts a prefix, suffix, and "range to replace", allowing it to make complex changes all at once by rewriting the range To users, this appears as an inline diff

171

Nate Sesti @natesesti.bsky.social · 04/09/2025

Our goal in training and open-sourcing Instinct is to lay a foundation for even stronger open models and enable self-hosting. Check out our blog for full details: blog.continue.dev/instinct

blog.continue.dev

Introducing Instinct: the world’s best open Next Edit model, built by Continue

Meet Instinct, Continue’s open Next Edit model that's built to predict your next edit and keep you in flow

120

Nate Sesti @natesesti.bsky.social · 04/09/2025

Introducing Instinct 🐺, a new state-of-the-art open Next Edit model trained by @continue.dev Runnable with @ollamabot.bsky.social and available today in VS Code

1164