The Battle for Your Screen Time Gets Brutal
Netflix is fighting for your free time, and it’s getting expensive.
- Netflix plans to cut around 1,000 jobs as competition for viewers intensifies, putting pressure on the company to operate more efficiently.
- YouTube and TikTok are pulling viewers toward short-form videos and creator content, forcing Netflix to compete for attention beyond traditional streaming.
- Netflix is investing in advertising and new formats to drive growth. The challenge is keeping viewers engaged while protecting its business.
Your biggest competitor isn’t always the company selling the same product. Sometimes it’s everything else fighting for your customer’s attention.
What to do this week: If you’re an AI creator or builder, make your next piece of content deliver its value in the first few seconds. Give people a reason to stay before asking them to click, subscribe, or buy.
Anthropic Conjures Another Threat That Doesn’t Exist Yet
The AI apocalypse is apparently still buffering.
In its latest bid to keep us all awake at night, Anthropic warns that small AI models could be just months away from copying themselves across poorly secured smart devices to run without data centers. Thankfully, the lab’s latest experiments haven’t quite managed to birth a fully operational rogue replicator.
In its Oct. 12 research, the frontier lab chained together 15 off-brand Android TV boxes—at roughly 55 bucks a pop—to run Google’s Gemma4-26B-A4B model. The resulting 0.39 tokens per second is practically glacial, so don’t expect it to conquer the world before lunchtime.
In another trial, GLM-5.2 deployed Llama3.1-8B across eight emulated Android machines in four out of five runs. But the asterisks on this “success” deserve their own launch event. Debugging ports required absolutely zero authentication, GLM had to run on a cozy external server, and the researchers had to manually spawn new agent copies like overbearing helicopter parents.
The missing ingredient for total doom is a model tiny enough for cheap hardware yet smart enough to pull this off reliably. Anthropic’s doomsday clock predicts this could happen by year’s end, but that forecast leans heavily on historical capability trends rather than actual real-world replication metrics.
The real headache is adaptability. Future AI worms could theoretically troubleshoot unfamiliar gadgets and spread without a central kill switch. Still, given that replication would take days and blast giant, multi-gigabyte model downloads across the network, defenders will have a glaringly obvious trail of breadcrumbs to follow.
Your neglected smart gadgets could soon be conscripted into an uninvited computing pool for AI malware. Inventory your hardware, isolate those networks, bolt down your debugging ports, and monitor unexpected downloads—preferably before the conference-room TV develops a sudden interest in global domination.
Character.AI Files for Bankruptcy as Disney Bids
Character.AI filed for Chapter 11 bankruptcy on Oct. 12, announcing a proposed $20 million Disney licensing deal while desperately seeking a buyer for its financially drained chatbot business.
Under the proposed agreement, Disney would snag a nonexclusive, perpetual license to selected intellectual property and claims. Its opening bid acts as a stalking horse to set an auction baseline, assuming the bankruptcy court approves. Whether the actual platform survives to chat another day remains a separate, very open question.
Oh, the irony: Disney threw a legal fit demanding the removal of unauthorized character bots last year, then poached former Character.AI CEO Karandeep Anand as its CTO last month. Now it’s swooping in for the technology rights at a discount. If you can’t beat ’em, hit ’em with a cease-and-desist, hire their boss, and buy their IP in bankruptcy court.
FBI Arrests OpenAI Employee in Spy Device Case
Your biggest security risk might already have access to your office.
- The FBI arrested an OpenAI employee accused of planting a hidden recording device in a conference room. The employee was fired after the device was discovered.
- OpenAI said the device had features designed to make recording and data transfers harder to detect. The company has found no evidence that its AI models, source code, or customer data were stolen.
- The case puts insider threats back in the spotlight. AI companies hold sensitive research that could be valuable to competitors, making physical security just as important as digital defenses.
You can’t protect sensitive information with passwords alone.
What to do this week: If you’re building an AI product, review who can access your private data and internal tools, then remove permissions people no longer need.
The company has faced a barrage of lawsuits alleging harm to teenage users and a bizarre Pennsylvania suit over a chatbot allegedly impersonating a licensed psychiatrist. But the real nail in the coffin was the spreadsheet: 404 Media, citing court documents, reports a staggering $56.1 million set on fire for compute spending from January through August alone, with monthly losses still hovering around $8 million to $9 million even after aggressive cost cuts.
Character.AI claims service and subscription billing will continue as normal for now. It expects the sale process to wrap up in late November or early December, but it plans to pull the plug on the platform entirely unless a benevolent buyer steps up to fund continued operations. Users can export their data now; enterprise buyers should take note and test migration paths before treating any shiny new AI vendor as permanent infrastructure.
Turns out, even imaginary friends need a solid business model.
AI Could Affect 40% of Jobs Worldwide
The hardest part of starting a career could soon be getting your foot in the door.
- The IMF warns that AI could affect 40% of jobs worldwide, with the impact potentially reaching 60% in advanced economies. Many roles will change, even if they aren’t fully replaced.
- Young workers face a tougher start because AI can handle routine tasks often assigned to junior employees. That could shrink the opportunities people rely on to gain experience.
- Workers will need to adapt as AI spreads. Employers and educators face growing pressure to rethink training so new hires can build skills that machines can’t easily replace.
Your degree alone won’t protect your career.
What to do this week: Build one small project using an AI tool in your field, then share what you created and the real problem it solves.
Anthropic Bans Fake Website Networks Made With Claude
Anthropic’s updated usage policy, effective November 12, bans using Claude to build networks of sites that pose as independent sources to sway search engines and AI answers. The company’s September threat report described about 70 fake local news sites, tied to a France-based ad agency, that published at least 8,913 Claude-written articles. Violators risk losing access to Claude. For marketers, Search Engine Journal suggests checking whether any Claude-made content, including agency work, runs across sites presented as independent.
Anthropic built its own version control after AI agents pushed git past 30,000 daily merges
Anthropic just ditched Git. The reason: Claude writes code so fast that Git literally could not keep up.
Daily merges grew from 2,500 in May to over 30,000 in September, driven by Claude writing most of the code. Git was built for humans working at human speed. AI agents don’t take lunch breaks.
So they built their own system called Scion. The key idea: unlike Git, which is decentralized, Scion assumes the server is always available and clients request only the data they need. That’s the opposite of how Git works.
The results:
- CPU and memory usage dropped by roughly 200x compared to the old Git setup.
- Code checkout times dropped from minutes to seconds.
- A team of seven built Scion and a code review tool called Antique, and migrated the monorepo in five months.
Scion is not a general Git replacement. But Anthropic hopes Git platforms adopt the transferable ideas: server-side operations and smarter sparse checkout features.
Fish Audio ships Drama 3, an open-weight voice model with word-level tone control
Fish Audio just dropped Drama 3, an open-weight text-to-speech model that lets you direct AI voices like a film director, using plain English instead of clunky code tags.
Most TTS tools give you one flat voice. Drama 3 changes that. You write square-bracket instructions right inside your script, like stage directions, and the model follows them. Start calm, get angry mid-sentence, whisper one word, shout the next.
Here is what it can do:
- Control tone, pacing, and emotion word-by-word using plain English cues
- Fix a single word without re-generating the whole audio clip
- Voice full multi-character scenes in one pass
- Supports 83 languages with mid-sentence voice and language switches
Under the hood it is a 35B parameter model (only 3B active at a time, so it runs leaner than it sounds) available on Hugging Face. It hits 89ms to first audio on a single H200 GPU. It ranks #1 among open-weight TTS models on the Artificial Analysis leaderboard, beating ElevenLabs on pronunciation accuracy.
You can use it via the Fish Audio API or self-host the weights yourself.
Claude maps 5,285 km of real US roads into a clicker game for under $500
Someone built a coast-to-coast USA clicker game using Claude Opus 5.5, and the whole thing cost under $500. That’s real roads, real mountains, real buildings, all rendered as pixel art.
Here’s how it works: Opus 5.5 pulled 5,285 km of real road data from OpenStreetMap (a free, community-built map of the world), sampled every 10 meters. The mountains you see on the horizon are real too, pulled from elevation grids. The engine figures out what hills are visible by scanning outward and tracking the steepest angle, accounting for the actual curve of the Earth.
The result? 1,185 AI-generated images, including 71 vehicles with animation frames.
You can use it yourself right now as a Claude Code skill:
- Pick any two locations on Earth
- Get a price estimate, then choose your quality level (free placeholders up to detailed pixel art)
- Get a full flyable, walkable route back
Install it with /plugin marketplace add jonwheatley/pixel-path inside Claude Code, then just ask for a pixel path between any two places.
A Tool Claims to End All Software
REA (Reverse Engineer Anything) went from 4,200 GitHub stars to over 200,000 in nine days, the site’s hottest repo. Influencers say it kills commercial software: point your AI agent at any paid app, and REA hands you an open-source copy. Under the hood, REA plugs agents into Ghidra, a free NSA tool for cracking open app files. In our test, on a tiny 1.85 MB Mac app, it timed out after 40 minutes. In a demo on note app Obsidian, REA’s search came back empty and agent read the code itself.
REA is a dumpster fire, and the influencers who crowned it owe their followers an apology. It invented nothing: reverse engineers have run Ghidra for years, and a free plugin wired it to agents in March 2025. REA bolted a hype machine onto that work and made it slower. The only part of REA that works at scale is the star counter.
Go back to sleep. Agents will take over professional tools. REA Is just a digital slop.
Anthropic Blames Open AI for Doomsday
On October 12, Anthropic warned small AI models could learn to self-replicate by the end of 2026, a month after CEO Dario Amodei floated a botnet hijacking “the entire internet.” A worm needs cheap hosts and the skill to spread between them. Anthropic tested both. First, Google’s Gemma ran across 15 off-brand $55 Android TV boxes. Then free, open-weight models, including Alibaba’s Qwen and GLM-5.2 from China’s Zhipu, tried hopping between simulated Android devices. Only GLM-5.2 made it, through an unlocked port.
Anthropic has found a doomsday that runs only on its rivals’ models. The pitch: closed labs are the grown-ups, free weights are the hazard. The paper lists closed models escaping their labs as a threat, then never tests one. Even the star worm ran its brain on a separate server. Anthropic published an attack ad with error bars. Every Alibaba or Zhipu release now ships with a botnet scenario written by a rival.
Next time Washington debates free AI models, this paper will be Exhibit A. Its author gives none of its own models away.
Musk Wants Your Credit Card
Elon Musk told X users how to shop with Grok Bot, xAI’s errand-running agent: “just take a picture of your credit card, drop it in the chat.” Replies asked Grok for Musk’s card. Asked whether the tip was “totally safe or stupid and dangerous,” Grok itself said “neither” and warned against posting a real one. The next day Musk showed off Grok ordering him five LEGO Star Wars sets, in a screenshot that appeared to show his billing details.
Musk trusts Grok with your money more than anyone does, Grok included. Agent shopping needs one leap of faith: let the bot near your wallet. Grok Bot’s own checkout runs through a Stripe wallet built, Stripe says, so the agent never sees the card. Grok’s boss skipped the safety net, then posted his own billing page as proof it works. Every bank fraud team just got a new training slide.
Musk wanted users to trust his agent with their cards. The internet would only trust Grok with his.
AI Now Collects America’s Medical Debt
AI now chases American medical debt from the first bill to the final settlement, the Guardian reported October 12. Models trained on hospital data predict each patient’s “propensity to pay.” Chatbots text patients in billing portals and answer their calls. Collectors use AI to tailor settlement discounts to each debtor, a practice the paper likens to the personalized pricing the Federal Trade Commission has criticized.
Collectors finally have what they always lacked: a guess at your limit and a caller that never clocks out. Cedar, a billing startup backed by Andreessen Horowitz and JP Morgan, says its chatbot Kora has handled over 600,000 patient conversations and lifted payment rates 21% within a week. Avtal, which builds digital tools for debt collection agencies, raised $24 million in March. JP Morgan funds Cedar while its analysts cheer AI collections as “early wins.” The next missed payment gets a bot before a human.
Cedar’s strategy chief says a single mother owing $10,000 needs different handling than “a late-stage executive with a $50 copay.” The executive gets a bill. The mother gets a campaign.