The Daily Fledgling

The Daily Fledgling · The Daily Fledgling, October 30, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Fri, 30 Oct 2026 09:00:00 +0300

The Daily Fledgling · The Daily Fledgling, October 29, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Thu, 29 Oct 2026 09:00:00 +0300

The Daily Fledgling · The Daily Fledgling, October 28, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Wed, 28 Oct 2026 09:00:00 +0300

The Daily Fledgling · The Daily Fledgling, October 27, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Tue, 27 Oct 2026 09:00:00 +0300

The Daily Fledgling · The Daily Fledgling, October 26, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Mon, 26 Oct 2026 09:00:00 +0300

The Daily Fledgling · The Daily Fledgling, October 25, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Sun, 25 Oct 2026 09:00:00 +0300

The Daily Fledgling · The Daily Fledgling, October 24, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Sat, 24 Oct 2026 09:00:00 +0300

Grok Bot Moves Into Your Phone · The Daily Fledgling, October 23, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Fri, 23 Oct 2026 09:00:00 +0300

Grok Bot Moves Into Your Phone: Calendar, Reminders and Contacts on iOS and Android, sign-ins through 1Password, and a release that teaches every bot to make videos
Claude Stops Asking at Every Step: From next week, auto mode becomes the default in Claude’s web, desktop and mobile apps on Pro, Max and Team plans
Claude Writes 180 Million Lines of Proof: Anthropic and Theorem prove that 239 fuzzing entry points in 123 open-source libraries can never crash, checked on the machine code itself
Agent Skills Get a Neutral Home: Anthropic hands the SKILL.md standard to the Agentic AI Foundation, where MCP already lives

AI Finds a Hole Under Seventy Per Cent of the Web · The Daily Fledgling, October 22, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Thu, 22 Oct 2026 09:00:00 +0300

AI Finds a Hole Under Seventy Per Cent of the Web: PHP2Shell: two critical flaws in the PHP engine itself, found with frontier models, fixed in secret over two months and now patched
Claude Gets a Duet Partner: A proactive agent app that briefs you each morning, works on its own computer while you are away, and asks before it acts
OpenAI Moves Into Google’s House: GPT models and Codex workflows arrive in Google Cloud’s Model Garden
Leads in Mouse Brains and Busy Beavers: GPT-Rosalind points at cholesterol in a rare brain disease, and a Gemini pipeline finds 20 new mathematical “Cryptids”
Claude’s API Gets Its First New Version Since 2023: Temperature and friends are gone, and Claude Security now re-scans your code every time it changes
Inside ChatGPT’s Ad Counter: An LLM orchestrator guesses what you might need, an auction picks the ad, and the answer never waits for it
Doctors’ Files, Flattery and a Magic Square: A medical tie at the top, the oldest models flatter most, and four models hide a Gaudí puzzle in their 3D scenes
HeyGen Takes Both Top Voices: Voice Turbo leads the speech arena at $4 per million characters, Video 1 costs a cent a second, and Voxtral hears Gustave Eiffel
Grok Bot Learns to Make Pages: Reports, memos and dashboards you edit together, or let the bot keep up to date; someone already put Doom on one
Argon Eats 25 Trillion Tokens in Three Days: Plus a beating heart in 3D, and Sundar Pichai admits even Google cannot get enough compute
Muse Passes Ten Million in Forty Days: Zuckerberg talks up revenue per user, and Meta puts $1M behind an open-weights hackathon
Flint Sparks on a Laptop, Hermes Closes Jensen’s Show: A fast local coding model from Richard Socher’s lab, and Nous’s agent as the finale of NVIDIA’s Berlin keynote
Codex Grants Go Pro, a Physicist Joins a Trust & IKEA Goes 3D: Plus Mistral and NVIDIA in the factory, Karpathy on strange chains of thought, and a French fork of Battle Drones

Claude Designs Binders Against Scorpion Venom · The Daily Fledgling, October 21, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Wed, 21 Oct 2026 09:00:00 +0300

Claude Designs Binders Against Scorpion Venom: In 64 hours it drew up 318,484 protein designs; labs tested 211, and 17 latched onto toxins from Mexico, North Africa and the Middle East
DeepSeek Opens the Engine Room: DeepInfer, the in-house framework that serves V4.1-Flash, goes fully open source, with the API’s token handling alongside
Mistral’s Voxtral 3 Learns to Listen, Think and Talk: New transcription and speech models, a 30-minute voice clone and an agent kit that answers in under a second
OpenAI Buys Orchestra Bio, and Madry Comes Back: A drug-development platform joins the Rosalind life-sciences push, and a departed researcher returns to work on superintelligence safety
Day 16: The Cache That Never Forgets: Twelve-hour prompt caching in the API, a forever cache in Codex, and credits for 40,000 accounts that were billed on the wrong limits
ChatGPT Gets a Locked Drawer: Locked Chats hide sensitive conversations behind a PIN or Face ID, with their own memory
Yesterday’s Genius, 86 Times Cheaper: Artificial Analysis tracks how fast each level of intelligence gets cheap, and why the very best still costs more
Battle Drones, and a Periodic Table That Sings: Simon Willison builds a browser RTS with Claude where two LLMs can fight each other, and Gemini 4 Argon animates Tom Lehrer
Grok Bot Sweeps 204 Tickets in Ten Minutes: An incident sweep with one engineer approving the batch, plus a long list of new features
Muse Wonder Spreads, MiniMax Wins on Thrift & Ant Group’s Tiny Voice: Plus Clef and Ornith on Ollama, Fibo 2.0 for images and EmbeddingGemma 2 on a Mac
Mastercard Pays Machines, PyTorch Leaves Home & Plants Get a Translator: Plus a de-bloated TV, Copilot Code’s timetable and a HeyGen voice clone

Anthropic’s Free Bug Hunt Reports 77 Critical Flaws · The Daily Fledgling, October 20, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Tue, 20 Oct 2026 09:00:00 +0300

Anthropic’s Free Bug Hunt Reports 77 Critical Flaws: OSS Scanner sends model-written reports, proofs of concept and patches straight to open-source maintainers, and curl has already fixed one of its worst bugs in years
OpenAI Teaches Its Models COBOL: GPT‑6.1 Sol migrates COBOL programs with 86.6% accuracy at low effort, up from 5.4%, and GPT‑6 Astra builds a SQLite clone in 20 languages
MiniMax M3.1 Flash: A 42 for Twenty-Two Cents: Artificial Analysis puts it 13 points above M3 at under half the cost per task, and Ollama hosts it free for three days
Meta Shows the Plumbing: Prometheus and Muse Glimmer: 85%+ goodput across hundreds of thousands of GPUs, a 30B open model for one consumer GPU, and Muse Wonder Full at No. 6 in WebDev
At Last, a Benchmark for Slop: AA‑PersonalityBench grades 30 personality traits and 21 habits of style, and finds Llama 4 Maverick the most padded and GPT‑6 Sol the leanest
Day 15: OpenAI Gets Cooked: A new show builds an 8‑bit ramen game, Codex cloud tasks keep running after the laptop closes, and OpenCode’s API users get refunds
Decks Get Co‑Authors in ChatGPT and NotebookLM: ChatGPT Slides adds live editing and comments; NotebookLM lets you review the outline before it builds a deck or a video
Grok Bot Moves to the Action Button: Version 0.71.2 brings bigger files and pre-filled forms, and Runway and HeyGen plug in
Exa Ties for First, FLUX 3 Max Rolls & HealthBench Gets a Table: A busy day on the scoreboards, from search APIs to text-to-speech
Unity Joins the Plugin Shelf, Cloudflare Shares Its Audit & Claude Catches a Miner: Plus OpenClaw’s sidebar, Copilot’s Autopilot, Wizard Chess in Argon and a plea to slow down

Anthropic Puts a Fifth of Itself Toward the Public Good · The Daily Fledgling, October 19, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Mon, 19 Oct 2026 09:00:00 +0300

Anthropic Puts a Fifth of Itself Toward the Public Good: A new fund starts with 5% of the shares and may grow by 2% for every 50% rise in value; with founder and staff pledges, roughly 20% is now committed
Day 14: Codex Lets You Edit the Past: The desktop app can now edit earlier messages in a thread; dots are coming to the EU; and Tibo gets asked for resets on the street
Altman on a Humanoid Robot: “We Will”: OpenAI promises “a cool demo in 2027” but no robots on the streets for years, and 27 robotics openings point to hardware made in-house
Claude Code Forks by Default, and Claude Design Gets a Closing Date: Version 2.1.302 turns on fork subagents everywhere, moves subagents to the background, and warns that standalone Claude Design closes by December 14
Epic and Spotify Hand the Agents New Instruments: An official Unreal Engine plugin puts hundreds of editor tools inside Claude Code; Save to Spotify turns agent-made audio into episodes in your library
An FPS, an Anime and a Game Boy Mario, All Made in Code: Builders spent the weekend showing what their agents made, and Thariq would now rather vibe-code games than play them
Glimpse Claims Sharp Eyes, pstack Proves Things & Grok Bot Passes 10,000 Ratings: A small visual-reasoning model; a formal-verification skill; two camps on agent UI; a free agents course and 177 science skills

DeepSeek’s Flash Gets Faster, Cheaper—and Stays Open · The Daily Fledgling, October 18, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Sun, 18 Oct 2026 09:00:00 +0300

DeepSeek’s Flash Gets Faster, Cheaper—and Stays Open: V4.1-Flash-1017 tops every model in its own table on DeepSWE, ships MIT weights, and arrives with a $0.75 search API
Day 13: Le Limit Reset: Tibo resets Codex limits again, is “working on two” of a user’s wishes—and lets two dots file his taxes
Muse Wonder Full Meets VulcanBench: Close to Astra, Opus 5.5 and Sol 6.1 on accuracy, but slower—and Alexandr Wang says Meta is on it
A 61 KB Ocean, a Mech Battle Royale and a 42-Bone Crab: The weekend’s Claude Code show-and-tell, plus an official plugin that prints a receipt for your usage
OpenCode Discloses a Breach of Saved Provider Keys: A “large-scale distillation attack” hit 1.4% of hosted Console users; some report charges they never made
Proving the Work Becomes the Bottleneck: Poteto teases a verification skill for pstack as Addy Osmani argues agents must check their own output
Meta Edits Its Terms, FLUX 3 Klein & Android Skills for Agents: Training clause dropped; a 5.6B open image model; 30 official skills from Google; OpenAI’s weekend-project gallery

Reflection Opens Beam, Its First Model · The Daily Fledgling, October 17, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Sat, 17 Oct 2026 09:00:00 +0300

Reflection Opens Beam, Its First Model: 501B total / 23B active under Apache 2.0—#2 among U.S. open weights, and free on Ollama for now
GPT-6 Astra Pro Learns to Split the Work: Several agents per task in ChatGPT’s Chat tab—and Codex stretches your usage with smarter caching
Four Argon Agents Slay the Ender Dragon: 27 hours, 2,783 self-written programs, no human help—while Argon spreads to AI Studio, Canvas and Perplexity
Claude Writes 90% of Claude Code. Here’s Who Checks It.: Merged PRs more than doubled while reverts fell 40%—and Claude Code now starts answering up to 3× sooner
Liquid’s d1-ultra Tops the Decision Index: A score of 65, earned by saying “I don’t know” 85% of the time it should
Claude Mythos 5 Draws Better Quantum Circuits: From plain-language instructions, designs that beat the best published ones—sometimes by an order of magnitude
RTX Spark Laptops Ship With Agents Inside: OpenClaw and Hermes Agent set up from Windows, no terminal required
Alexandr Wang: “We Have Built a Machine That Builds New Models”: A year into Meta Superintelligence Labs, a thank-you note for reaching the frontier—with no deadlines
Jalapeño at TIME, Copilot Notebooks & a Navier-Stokes Shirt: OpenAI’s chip honored; Office editing in Notebooks; Mistral on STEMBench; HeyGen Video 1 lands in MCP

Meta Opens the Weights on Muse Spark · The Daily Fledgling, October 16, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Fri, 16 Oct 2026 09:00:00 +0300

Meta Opens the Weights on Muse Spark: 816B total / 42B active, 1M-token context—and Artificial Analysis crowns it the top U.S. open-weights model
Gemini 4 Argon Lands in Antigravity for Pro: A separate weekly Argon quota—so Flash stays free for the rest of the week’s work
GPT-Live-1.1 Speaks Up; ChatGPT Drives Chrome: New language voices across ChatGPT, Codex and dots—plus computer use for anyone with the Chrome extension
Google Cultivates a Magic State on Willow: First hardware demo of magic-state cultivation—40× fewer errors, 99.99% logical fidelity, in Science today
Opus 5.5 Fast Mode Exits Preview; Code Review Goes GA: Same Opus, up to 2.5× faster at $8/$40 per million—and Team/Enterprise get Quick, Standard or Deep PR reviews
Day 10 Reset, Day 11 Capacity—Codex Keeps Shipping: Tibo’s reset button returns; capacity errors should self-heal—and the Codex app gains composer predictions
Simons Puts $15M on AI-for-Math; Anthropic Opens the Lab Door: OpenAI and Anthropic each pledge $5M; METR gets employee-like access; two new Anthropic fellowships open
Mistral and CERN Chase a Large Physics Model: One model on billions of simulated collisions—for reconstruction, ID, anomalies and future detectors
TIME Inventions, Rosalind Labs, Gemma Docs & Hermes: SynthID and Omnilingual ASR honored; biology access expands; NVIDIA models land in Hermes

Claude Code Desktop Learns to Juggle Sessions · The Daily Fledgling, October 15, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Thu, 15 Oct 2026 09:00:00 +0300

Claude Code Desktop Learns to Juggle Sessions: Nested agents, a live to-do rail, /btw side tabs—and a v2-feel redesign for parallel work
Grok Voice 3.0 Takes the Speech Crown: #1 on Artificial Analysis at 86.9—and SpaceX says it already closes 70% of Starlink support calls
Argon Clocked In at the Lab Bench: Enzymes in hours, a 60-year plasma conjecture challenged, and 10,000 scientists get free Ultra Antigravity
ChatGPT Space: Shared Docs That Stay Alive: DevDay’s team room—turn a messy chat into a living plan your coworker can edit with the model
Hermes Agent Crosses the Threshold: v0.22.0 ships—and 50+ connectors go live with no account required
Singapore, Design, Launchers & Robot Arms: Anthropic’s fifth APAC office, Mistral buys New Material, Epic vibe-codes a launcher, and AA grades robot models

Gemini 4 Argon Rolls Out Globally · The Daily Fledgling, October 14, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Wed, 14 Oct 2026 09:00:00 +0300

Gemini 4 Argon Rolls Out Globally: Google’s new frontier model for hard real-world work—and it taught a robot cornhole from a video
OpenAI Finds a Bug in Its Own Office: A planted recorder, models that traced it, an FBI arrest—and no sign of stolen weights
Anthropic Ditches Git for Agent Scale: From 2,500 merges a day to 30,000—so they built in-house source control in five months
XBOW Banks a $250,000 Chrome Bounty: Second full-chain award in Chrome history—built with OpenAI’s GPT-5.6 Sol in six days
DigUp Makes Spotlight Understand Your Files: A free macOS menu-bar app on EmbeddingGemma 2—search photos, video, audio and code by meaning, fully local
Potholes, Private Apps, a Chapter 11 & a Disk Diet: Claude vs Amsterdam asphalt, vibe apps on Play, Character.AI’s bankruptcy rumor mill, and Codex’s 60% thinner desktop

Meta Unveils Muse Wonder · The Daily Fledgling, October 13, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Tue, 13 Oct 2026 09:00:00 +0300

Meta Unveils Muse Wonder: Zuckerberg’s new frontier family for personal and professional agents, at $4/$20 per million tokens
Anthropic Says Distillation Attacks Are Down Over 90%: A research note on how Claude is being hardened against illicit copying of frontier capabilities
Grok, Cursor and X Premium Share One Bill: SpaceX Super Intelligence unveils “The X Subscription,” five tiers from Free to Ultra
Day 8 of the Sprint: A Global Reset Lands: OpenAI’s Tibo posts “Day 8/ Reset,” then confirms the usage refresh is done
EmbeddingGemma 2 Turns Sketching into a Game: Google Gemma’s unified embedding space for text, images, video, audio and code
Ambassadors, Sandbox Warm-ups, a Lobster Patch & a Robotaxi Bot: OpenAI Dev news, OpenClaw 2026.10.2, and Grok Bot rides in a Cybercab

OpenClaw Issues Version 2026.10.1 · The Daily Fledgling, October 12, 2026

noreply@thedailyfledgling.click (Nadav Solomon) — Mon, 12 Oct 2026 09:00:00 +0300

OpenClaw Issues Version 2026.10.1: Skill discovery, MCP Apps beside the chat, and full backups to Cloudflare R2
Karpathy: The Gap in Understanding Has Become a Funnel: Billions have barely met an LLM; a few thousand see the frontier from inside
Hermes Agent Gets Eyes on Your Browser: A community-built, open-source side panel feeds the page at hand to Nous Research’s agent
OpenHands’ “Agent Canvas” in the Spotlight: One self-hosted control room for Claude Code, Codex, Gemini and more
“Anatomy” Joins Hermes Agent’s Skill Shelf: An open-source skill that explains ideas by building working machines
“We Shouldn’t Settle for an Opaque Superintelligence”: Thariq sides with Terence Tao on understanding over mere results