---
title: 'In the News: October 1, 2026 (Evening)'
description: 'Earendil ships Pi Durable, an experimental harness that checkpoints every model call and tool call so agents resume after a crash, alongside Pi 1.0.'
canonical_url: 'https://darkfactory.dev/news/2026-10-01-evening'
markdown_url: 'https://darkfactory.dev/news/2026-10-01-evening.md'
collection: news
date_published: '2026-10-01T19:00:00-04:00'
date_modified: '2026-10-01T19:00:00-04:00'
---

# In the News: October 1, 2026 (Evening)


Earendil released Pi Durable today, an experimental package that treats every model request and tool call as a checkpointed task, so an agent whose process dies can resume from the last checkpoint. It shipped the same day as Pi 1.0, the stable release of the Pi coding agent.

## 1. Pi Durable makes crash recovery a harness primitive

**[Pi Durable](https://earendil.com/posts/pi-durable/)** · Earendil Engineering (Earendil Inc.) · earendil.com, October 1, 2026

Earendil defines a harness as "storage plus the machinery needed to run one or more conversations with large language models in parallel." In Pi Durable, "everything the harness runs, from calling the model to executing a tool, is a task," and each task stores a checkpoint before it moves on. After a crash, a new process opens the same storage and continues unfinished tasks. A cut-off model request is sent again, with the partial answer kept in the transcript and marked as aborted. A cut-off tool call reruns only if the tool declares itself safe to replay; otherwise the model is told the call was interrupted. A `requestId` makes a submission exactly-once.

The package also covers concurrent conversations that fork from a point in a transcript, background compaction that runs while the conversation continues, and extensions that can be swapped on a running agent. Earendil says the source is about 15,000 lines without tests, roughly 150,000 tokens with GPT and 250,000 with Claude, and that it ships memory, SQLite and JSONL storage. These are Earendil's own figures; we have not run the package or checked them. The post calls Pi Durable experimental and says the API might still change.

Armin Ronacher (Earendil) and Mario Zechner answered questions in the Hacker News thread, which showed 130 points and 12 comments at an early check about two hours after posting.

**Why it matters:** The replay declaration is the part worth copying: a tool that deploys is reported to the model as interrupted, never repeated, while a read-only search reruns. If your agents run unattended for hours, decide per tool what a restart does before the first crash. Treat the package as a reference design until the API settles.

## 2. Pi 1.0 adds deferred tool loading and mid-conversation system messages

**[Pi 1.0](https://earendil.com/posts/pi-1-0/)** · Earendil (Earendil Inc.) · earendil.com, October 1, 2026

Earendil lists what 1.0 adds: Codemode with native MCP support, extension support for virtual models, deferred tool loading, cache warming for Anthropic models, mid-conversation system messages ("transcript-aware prompt and tool changes"), a new TUI theme, and full-screen mode by default. The post says "hundreds of thousands of people" use Pi weekly, a figure we could not verify. It also says the list of features that were tried and dropped "is much longer" than the list that shipped.

On Hacker News the release thread showed 473 points and 164 comments at an early check about two hours after posting. Ronacher described how they choose what to build: "we look at what the models are doing. They are trained on their respective harnesses and we're not here to fight their behavior."

**Why it matters:** Deferred tool loading and prompt changes that leave the cache valid both target the cost of a large tool set in a long session. The post gives no measurements of either, so test them on your own tool list before relying on them.