Frontier Daily (Sep 19): 17 items from One Useful Thing, Hacker News, Simon Willison and more + X supplements

Every day Laojin pulls the official channels and compiles the past 48 hours of AI news into a short brief: curation only, no commentary, with the original link and publish time attached to every item.

1. The Overhang

The post argues that AI is still on an exponential curve while human systems lag behind, and that focusing only on future AI overlooks how capable current models like GPT-6 Astra and Fable 5.1 already are.

Source: One Useful Thing · 2026-09-18 · Original

2. US Military had close call after using AI for hallucinated intelligence report

A Hacker News post with 143 upvotes and 69 comments reports that the US military had a close call after using an AI-generated intelligence report that turned out to be a hallucination.

Source: Hacker News · 2026-09-18 · Original

3. AI is an elite crime spree

An article titled "AI is an elite crime spree" drew 109 upvotes and 35 comments on Hacker News.

Source: Hacker News · 2026-09-18 · Original

4. The Creative Spirit of Who Framed Roger Rabbit

Simon Willison revisits Robert Zemeckis's 1988 film Who Framed Roger Rabbit, highlighting a scene of a pelican riding a bicycle where the pelican was animated and the bicycle was real, reportedly stabilized with water-filled wheels and guided by a cable.

Source: Simon Willison's Weblog · 2026-09-18 · Original

5. modelscope/sirchmunk: 🐿️ Sirchmunk: Raw data to self-evolving intelligence, real-time.

ModelScope released Sirchmunk on GitHub, a Python project aimed at turning raw data into self-evolving intelligence in real time, with 1,318 stars.

Source: GitHub (ModelScope) · 2026-09-18 · Original

6. [AINews] not much happened today

AINews recaps September 16-17 developments, noting Anthropic's Projects in Claude Code enabling parallel cloud sessions from one conversation, and Google's Gemini managed agents update with an Antigravity-based harness plus Credentials and Files APIs.

Source: Latent Space · 2026-09-18 · Original

7. Be alert: targeted attacks on prominent Rustaceans

Adam Harvey and the crates security team warn of an ongoing campaign targeting rust-lang members and popular crate owners, using video calls to trick targets into installing malware or running commands, a technique used last month in a successful supply chain attack on the arrayref crate.

Source: Simon Willison's Weblog · 2026-09-17 · Original

8. How Cooley is accelerating IPO work with ChatGPT

Cooley built GO Public with ChatGPT Work to bring intelligence into the IPO process, helping lawyers surface issues earlier and focus judgment where it matters most.

Source: OpenAI Blog · 2026-09-17 · Original

9. Introducing Astra for Law

OpenAI for Law brings frontier intelligence for law, custom firm workflows, connected legal data sources, and legal-grade controls for confidential client work.

Source: OpenAI Blog · 2026-09-17 · Original

10. OpenAI Launches Astra for Law with GPT-6 Astra

OpenAI announced Astra for Law, a frontier intelligence offering powered by GPT-6 Astra with a new Legal Search Index, specialized settings, tools, and dozens of partner/community plugins tailored for legal workflows.

Source: X (Grok x_search, @OpenAI) · within 48h · Original

11. Anthropic Optimizes Biomolecular Models and Launches Protein Design Competition

Anthropic showed how Claude optimized inference for over 30 open-source biomolecular models (4x faster on average) and open-sourced the code, while partnering with Adaptyv Bio on a protein design competition with up to $1M in credits and validation funding.

Source: X (Grok x_search, @AnthropicAI) · within 48h · Original

12. Google DeepMind Releases AlphaGenome Atlas

Google DeepMind launched the freely accessible AlphaGenome Atlas, a predictive map of every possible DNA letter change in the human genome, already aiding researchers in identifying disease variants and regulatory patterns with improved rare genetic signal detection.

Source: X (Grok x_search, @GoogleDeepMind) · within 48h · Original

13. Alibaba Qwen Releases First Omni-Modal Model Qwen3.8-Omni-Flash

Alibaba's Qwen team released Qwen3.8-Omni-Flash, their first omni-modal model with native audio-video understanding, reasoning, tool use, and 1M context; it approaches Gemini-level performance in agentic tasks while cutting video input costs by ~89% and open-sourcing plugins.

Source: X (Grok x_search, @Alibaba_Qwen) · within 48h · Original

14. Mistral AI Denies Unauthorized System Access

Mistral AI stated that after a thorough investigation, it found no evidence of unauthorized access to its systems and confirmed that its systems have not been compromised.

Source: X (Grok x_search, @MistralAI) · within 48h · Original

15. Anthropic Proposes New Metrics for AI Development Transparency

Anthropic shared three internal measurements (AI's share of R&D, quality of AI agent oversight, and compute allocation) to track frontier progress and called on other labs to publish similar metrics for better public understanding.

Source: X (Grok x_search, @AnthropicAI) · within 48h · Original

16. Elon Musk: AI Will Roughly Double US GDP Growth

Elon Musk guessed that AI will roughly double US GDP growth next year from ~2% to ~4% or more, while also highlighting that Grok Bot now has voice capabilities.

Source: X (Grok x_search, @elonmusk) · within 48h · Original

17. DeepSeek Releases Efficient V4.1-Flash Multimodal Model

DeepSeek introduced V4.1-Flash, a 552B MoE model with asymmetric Causal Encoder-Decoder architecture (8B active params for input), drastically reduced KV cache, lower pricing, and native multimodal support that outperforms prior versions on benchmarks and cost.

Source: X (Grok x_search, @deepseek_ai) · within 48h · Original

---

_Sources: official blogs of OpenAI, Google DeepMind and Qwen; official RSS of Simon Willison, Latent Space, Import AI and One Useful Thing; Anthropic site monitoring, top AI discussions on Hacker News and open-source updates from key AI orgs on GitHub. X sources are retrieved live via Grok x_search with per-link verification. This is a source digest — no commentary from Laojin._