AINews：Pi 1.0 与 Pi Durable 同登 HN 首页——codemode、虚拟模型扩展，TS 移植版把 agent 状态全外置做"崩溃可续跑"

正文

Last call for regular tickets for AI Engineer NYC ! See you in 2 weeks!

As an exclusive for Latent Space subscribers, the first 30 of you can take a 30% off code if it helps (for new tickets only, no refunds).

Pi is often mentioned in the same breath as OpenClaw, as we did earlier this year :

but today is time for the increasingly well regarded Earendil , which Pi joined, to have its day in the sun, with both Pi 1.0 and Pi Durable hitting the front page of HN.

• Codemode (native support for MCP, Jev and image models)
• 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
• Full-screen mode by default

Pi Durable ports Pi to TypeScript and externalizes all stateful components of Pi:

• Crash Survival: Every step is recorded as a checkpointed task. If a process fails or restarts, agents and subagents automatically resume from their last exact state.
• Portability: It runs anywhere with a JavaScript runtime (like Node, Bun, or Cloudflare) and uses pluggable storage backends (Memory, SQLite, JSONL) and flexible remote or local execution environments.
• Concurrency: A single harness can run multiple parallel, branching conversations—such as a main channel and separate threads—without blocking one another.
• Extensibility: Developers can bundle custom system prompts, tools, hooks, and durable tasks (e.g., multi-step checkout processes with rollback capabilities) into installable “Extensions.”
• Context Management: Automatic background compaction summarizes older messages to maintain token limits without pausing the agent’s active work.
• Multiplayer & State Sync: Application state (like a to-do list) is stored in documents directly alongside the conversation transcripts, allowing multiple users or UIs to connect, watch, and steer the same agent simultaneously.
• Hot-Swapping: Tool and extension code can be updated dynamically while the agent is running, with the next tool call automatically picking up the new code.

AI News for 10/01/2026-9/30/2026. We checked 12 subreddits, 544 Twitters and no further Discords. AINews’ website lets you search all past issues. As a reminder, AINews is now a section of Latent Space . You can opt in/out of email frequencies!

Frontier and Multimodal Launches: Gemini 4 Argon, GPT-6.1 Sol and FLUX 3

• Gemini 4 Argon : Google announced a new generation of Gemini, with contributors highlighting revised pretraining mixtures, long-horizon post-training data, and internal applications in memory optimization, code migration and mathematics. These are developer accounts of how the model was built and used—not independent evidence of general superiority ( Google researcher ). Validation : Google says new Gemini revisions now undergo weeks of testing by thousands of internal software engineers before release ( Logan Kilpatrick ).
• Contested readiness : A circulated Bloomberg report attributed coding weaknesses to anonymous insiders; a subsequent post reported a senior DeepMind engineer rejecting that account. Treat the practical coding-quality dispute as unresolved, rather than interpreting either benchmarks or employee reactions as decisive ( reported criticism , reported rebuttal ).
• GPT-6.1 Sol : OpenAI’s update is primarily an efficiency story. Sam Altman called it the company’s fastest-growing model and said serving performance had improved after launch-time load problems ( update ). Measured economics : Artificial Analysis reports $0.72 per Intelligence Index task at maximum effort, versus $1.04 for GPT-6 Sol and $3.26 for Astra. Fewer turns and cheaper cache reads—not simply fewer generated tokens—drive the improvement ( results , explanation ).
• Multimodal fix : OpenAI also corrected image encoding for Luna and Sol. Luna gained one Intelligence Index point, including improvements on visual-document and knowledge-work evaluations; Sol changed negligibly ( measurement ).
• Solar Mini 4 : Upstage’s proprietary text-only reasoning model reports 35B total/3B active parameters, a 1M-token context window and 262K maximum output. Weights are not released, so parameter counts remain vendor-reported ( analysis ). Pricing : $0.10/$0.40/$0.01 per million input/output/cache-hit tokens.
• Trade-offs : Artificial Analysis scores it 24 overall and 83% on long-context reasoning, but only 1% on Terminal-Bench 4.0. Despite 208 tokens/s output, approximately 88K output tokens per task produce a 7.1-minute average completion time and roughly five times Luna’s task cost.
• FLUX 3 Image : Black Forest Labs launched native generation up to 4K, up to ten reference images, bounding-box layout control and targeted multi-turn editing. Preserving every untouched pixel is a vendor capability claim, not independently established here ( announcement ). Availability : Commercial weights are available; an open-weight variant is promised in coming weeks. Hosted access includes fal and Krea ( fal , Krea ).
• Pricing : BFL announced a temporary 50% API discount through October 8, without supplying base prices in these posts ( details ).
• Interactive video agents : Tavus introduced Griffin, a video-to-video interaction model. It claims 48% of live participants mistook it for a human, versus under 3% for earlier systems; that result should not be generalized into an unrestricted “Turing test passed” conclusion without the test protocol ( announcement ). Enterprise deployment : Separately, Synthesia launched Sessions: conversational avatars for roleplay and survey interviews, extending its previous one-way training-video product ( launch ).