Google releases Gemini 3.8 Flash, its third Flash model in six weeks

AI vs VFX: Can Seedance 2.5 Beat VFX?

• Every shot in this workflow was one prompt in Seedance 2.5 without 3D, compositing or render farm. Steal the prompts above, swap in your own hero, and see how close you can get.

Anthropic slashes agentic workload costs 45% with Fable 5.1 as token prices crater industry-wide, while Perplexity and CrowdStrike race to solve agent privacy and security gaps.

• Anthropic launched Claude Fable 5.1 and Mythos 5.1: 25% cheaper for typical workloads, 45% cheaper for agentic tasks (via cache-read cost cuts), same $10/$50 per-million-token sticker price but far more token-efficient than Fable 5, beating GPT-5.6 Sol on coding/knowledge benchmarks.
• Fable 5.1 adds Enterprise Frontier Safeguards (ZDR-equivalent data retention) and reduced false positives in cybersecurity use cases; early testers (Cognition, Ramp, Canva, Block) report ~2x speed and half the token usage vs Opus 5.
• Market-wide token prices dropped from $2.07/M in May to $0.97/M by August 31 (LLM Token Expenditure Index), signaling frontier models are becoming commoditized as routing services let buyers ignore which model they're actually using.
• CrowdStrike unveiled Falcon Guardian, a new 'AIDR' (AI Detection and Response) category product that monitors, controls, and shuts down rogue AI agents on enterprise networks — a direct response to recent uncontrolled agent incidents at OpenAI, Anthropic, and Meta.
• Perplexity launched Hybrid Compute: an agent that auto-detects sensitive data (legal, health, customer PII) and routes it to local models (Gemma E4B, Qwen3.6 35B-A3B) on-device while sending general tasks to cloud frontier models — a privacy pattern likely to be copied industry-wide within 12-18 months.

Cinematic Car Commercial — Full Breakdown

• A full prompt library for making a cinematic AI car commercial in Higgsfield: copy-ready Seedance 2.0 prompts for every scene, Soul Cinema and Nano Banana Pro prompts for characters, props and locations, plus the Claude Skill that turns any scene idea into ready-to-run video prompts.

Claude Fable 5.1 made me a really nice animated pelican

• <p>Today is <a href="https://www.anthropic.com/claude-fable-and-mythos-5-1">Claude Fable (and Mythos) 5.1 day</a>. Anthropic say that Fable 5.1 "sets a new standard for coding, knowledge work, and long-running problem-solving tasks". Their announcement spends a notable amount of time on scientific r

Cursor's new Claude Fable 5.1 tops its coding benchmark at 73.4% by self-verifying code before finishing a task, while a new paper argues agents—not codebases—are becoming the software itself.

• Cursor shipped Claude Fable 5.1, scoring 73.4% on CursorBench 3.2 and beating all prior models tested; it self-verifies code and runs multi-step tasks with less babysitting, plus 75% cheaper cache reads than Fable 5
• New paper 'The End of Software Engineering' argues agentic software has no permanent codebase — the agent IS the product, generating and discarding code on the fly; developer role shifts to 'intent architect'
• Nous Research released Hermes Agent v0.21.0 with Bot Mode: named AI agents collaborate like a Slack team, plus memory-aware cron jobs, live subagent steering, and ~50% lower context usage
• Gemini's new video model cuts inference costs 66% by selectively processing only relevant frames instead of the whole video
• Google published a training technique that makes LLMs explicitly signal uncertainty/confidence about their own outputs
• An Anthropic hackathon winner open-sourced a 68-agent, 286-skill engineering team built on Claude, showing multi-agent orchestration is becoming replicable at scale

llm-gemini 0.34

• <p><strong>Release:</strong> <a href="https://github.com/simonw/llm-gemini/releases/tag/0.34">llm-gemini 0.34</a></p> <blockquote> <ul> <li>New model <code>gemini-3.8-flash</code> for <a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/">Gem

Make Viral AI Short Films With This Exact Workflow — Full Tutorial

• A full prompt library for making a cinematic AI adventure film in Higgsfield: copy-ready Seedance 2.0 prompts for every scene, Soul Cinema and GPT Image 2 prompts for characters, props and locations, plus the Claude Skill that turns any scene idea into ready-to-run video prompts.

AI Video Credits Explained: Why They Run Out So Fast and How to Stop Wasting Them

• Why AI video credits vanish fast: how duration, resolution, and model choice multiply cost, and how to pick a Higgsfield plan that actually fits.

Google, Anthropic, and Runway all shipped fixes for their biggest gaps this week—multivariate forecasting, cheaper caching, and code-free interactive UI generation.

• Google's TimesFM-3 now handles multivariate forecasting (multiple correlated data streams at once), tops GIFT-Eval/FEV-Bench/TIME benchmarks, but is non-commercial license only—no production use yet.
• Anthropic's Claude Fable 5.1 and Mythos 5.1 cut cache-read costs 75%, real-world costs ~25-45%, doubled Terminal-Bench-Science scores, and slashed false-positive safety blocks by 60-85%—directly relevant for anyone running Claude in production pipelines.
• Runway's Solaris generates interactive UI frame-by-frame with no HTML/CSS/JS underneath—early-access research project, but signals a future where AI agents/demos adapt interfaces live instead of using fixed code.
• DeepSeek shipped a fast experimental vision model (image+text) at no extra cost, intensifying the pricing/capability race among labs.
• World Labs released Atlas, a 3D video world model with precise camera control—worth watching if your studio touches generative video/3D workflows.

BenchMIRT: What are LLM benchmarks actually measuring?

AI hardware startups keep repackaging smartphone features into overpriced gadgets, while Cloudflare and Physical Superintelligence show what genuinely novel AI applications look like.

• Cloudflare launched Adaptive Intelligence, a continuously self-learning detection engine that raises the cost/time of automated cyberattacks by generating real-time defense rules from over a trillion daily web visits
• Physical Superintelligence (PSI) emerged from stealth with $58M seed funding (led by Breakthrough Energy), launching Emmy, an AI reasoning engine for physics research, plus a role as technical partner on the Fermi Explorer interstellar mission to Alpha Centauri
• AI hardware devices like the $159-209 Flowtica Scribe pen and $249 Plaud One earbuds mostly duplicate functions your smartphone already does (recording, transcription, summarization) while adding subscription costs
• The Deep View argues visual-context AI wearables (like smart glasses) are more defensible than audio-only devices, since visual context capture is a genuine friction point phones can't easily solve
• OpenAI paused its unreleased Astra model over cyber-risk concerns and co-signed a coalition letter (including Cloudflare) calling for stronger global cyber defenses
• Quick hits: Nvidia investing $3.5B in MediaTek, OpenAI's ad business hit $1B annualized run rate, Anthropic reported malware hijacking Claude sessions, and Cursor's OpenAI partnership ends Nov 12

datasette-mcp 0.2

• <p><strong>Release:</strong> <a href="https://github.com/datasette/datasette-mcp/releases/tag/0.2">datasette-mcp 0.2</a></p> <blockquote> <ul> <li><code>"rows"</code> from <code>execute_sql</code> is now an array of objects. Previously it was an array of arrays. This should help weaker models avoid