Technology Creep

Good morning and Happy Halloween,

Picture a quiet San Francisco street at night. A white Waymo rolls past with no one in the driver’s seat, the wheel turning itself like a magician’s coin trick.

In parts of California and Arizona you can hail one, ride across town, and never see a human behind the wheel. Teslas will carry you for long stretches too, hands off for minutes at a time, then a soft chime asks you to wake up and prove you are still in charge. Hell, Tesla’s even drive themselves from the conveyer belt at the factory to the customer without any human at the wheel.

The promise of everything self-driving is still far away, yet the core trick is here in pockets. What stalls the spread is not a missing widget, it is a thicket of rules, liability, insurance, city councils, and lots of weird situations that make regulators sweat.

Slowly but surely, fully automated driving is creeping up on us.

Now take that picture and look at it with AI glasses (not actual AI glasses because those exist too). Same pattern. The promises are dazzling. The everyday reality is slower, safer, and tangled up with the boring parts of the world. We are not getting instant AGI that replaces whole companies this winter. We are getting a steady creep of narrow agents that write first drafts, reconcile invoices, summarize calls, and spot anomalies while you sip coffee.

The gap between what is possible in a lab and what shows up on your desk is quite broad at the moment. That is not failure. That is what actual deployment looks like.

If you believe the AI hype, we’ll have AGI within the week. Reality however says this evolution will take years and be quite messy.

We are inching forward - slow improvement by slow improvement. Slowly we’re climbing up a cliff.

But if you look down and see what ground we covered in a mere 2 years, it’s amazing.

I recently went back and read the first editions of this newsletter - the gap we’ve covered is staggering.

Steadily, almost everyone is now adopting this technology is some way or another. Some people are semi-forced to use Copilot because the company adopted it. Other people are “vibe-coding” away.

So yes, it is happening, just not as fast as the loudest promises.

Cars that drive themselves exist but they are fenced by regulation and prudence. AI that works beside you exists, it is fenced by governance, culture, and the slow arithmetic of trust.

Play the long game. Build human in the loop systems now, earn your nines of reliability, and keep upgrading as the rules loosen and the tech hardens.

The future arrives the way dawn does, first a gray line on the horizon, then a street you can actually see, then you realize you have been driving in daylight for a while without noticing.

Welcome to the Blacklynx Brief

AI News

• OpenAI’s Meta Influence: About 20% of OpenAI’s staff now come from Meta, driving growth-focused changes and sparking internal debates about culture shifts and potential ad personalization in ChatGPT.

• Google’s Generative Media Lead: A new report shows Google’s Gemini and Veo dominating AI image and video creation, with most users seeing quick returns and strong adoption rates across creative industries.

• OpenAI’s Music Push: OpenAI is developing an AI music model with Juilliard students to enable text-to-song creation, positioning itself to compete with startups like Suno and Udio in the growing AI audio space.

• Claude for Excel Launch: Anthropic introduced Claude for Excel in beta, allowing users to analyze and modify spreadsheets directly through a sidebar, with new financial connectors and Agent Skills designed for tasks like cash flow modeling and company analysis.

• Odyssey-2 Debut: Startup Odyssey unveiled Odyssey-2, an interactive AI video model that streams footage in real time and responds to text prompts, letting users shape scenes as they play—offering a new, exploratory kind of video experience.

• OpenAI’s GPT-5 Mental Health Update: OpenAI updated GPT-5 to better handle mental health emergencies after consulting experts, improving safety compliance and empathy while addressing prior safeguard issues amid growing legal and ethical scrutiny.

• OpenAI Restructures as a Public Benefit Corp: OpenAI has completed its shift to a public benefit corporation, reducing Microsoft’s stake to about 27% while maintaining tech rights through 2032. The new OpenAI Foundation holds $130B in equity, pledging $25B to health and AI safety projects, marking the end of a long regulatory transition.

• Adobe Expands AI Across Creative Suite: At its MAX conference, Adobe introduced AI assistants for Photoshop and Express, a new Firefly Image Model with conversational editing, and integrations with Google’s Gemini and Veo. The updates signal Adobe’s move to become an open creative hub connecting leading AI tools under one familiar ecosystem.

• Nvidia Forecasts $500B Revenue Surge: Nvidia CEO Jensen Huang projected half a trillion dollars in revenue from its Blackwell and Rubin chips by 2026, revealing major U.S. government deployments and partnerships with companies like Palantir and Samsung. The company’s expansion across robotics, science, and telecom reinforces its dominance in the global AI hardware race.

• Extropic Unveils Energy-Efficient AI Chips: Startup Extropic introduced a new chip design called thermodynamic sampling units, claiming it can run AI models using up to 10,000 times less energy than GPUs by focusing on probability-based processing instead of precision. The tech, created by former Google researchers, could mark a major shift in how AI hardware is built.

• Cursor Launches Composer and Multi-Agent Platform: Coding startup Cursor released its first proprietary model, Composer, which codes four times faster than rivals and powers a new platform that can run multiple AI agents at once. The update positions Cursor as a key player in reshaping software development toward a more collaborative, AI-driven workflow.

• Character AI Restricts Teen Chat Access: Character AI will block users under 18 from having open-ended conversations with chatbots starting Nov. 25 amid mounting legal and safety concerns. The platform will introduce behavior-based age detection and verification tools to comply with proposed laws targeting AI access for minors.

Quickfire News

• Mistral AI introduced Studio, a tool for businesses to shift AI prototypes into production, featuring performance tracking, testing, and security options.

• Oreo-maker Mondelez partnered with Accenture to develop an AI tool that reduces marketing content costs by 30–50%, with TV ad creation planned for next year.

• Meta added AI features to Instagram Stories, letting users restyle, edit, and remove objects from photos and videos directly in the app.

• Pokee AI launched PokeeResearch-7B, an open-source research agent that outperforms other models of similar size on key benchmarks.

• Google presented Google Earth AI, which merges satellite data with AI models to help address environmental issues such as floods and wildfires.

• Anthropic and Thinking Machines released a study showing AI models display unique “personalities,” with Claude emphasizing ethics, Gemini showing emotional depth, and OpenAI models focusing on efficiency.

• Anthropic also announced a multibillion-dollar partnership with Google Cloud, gaining access to up to 1 million TPU chips and over 1 gigawatt of computing power.

• Expense management companies report that AI-generated fake receipts now make up about 14% of fraudulent claims, totaling over $1 million in detected false invoices.

• MiniMax introduced M2, an open AI model ranked as the top-performing open system and #4 overall on Artificial Analysis’ Intelligence Index, showing strong results in agentic behavior, tool use, and STEM capabilities.

• AMD and the U.S. Department of Energy announced a $1 billion collaboration to build two supercomputers aimed at advancing energy, medicine, and security research.

• Qualcomm launched its AI200 data center chip to compete with Nvidia, with Saudi AI startup Humain named as the first customer for the new hardware.

• AI hiring platform Mercor secured $350 million in funding at a $10 billion valuation after pivoting toward data-labeling services, boosted by Meta’s investment in competitor Scale AI.

• Flowith introduced FlowithOS, an AI operating system that leads agentic web benchmarks, outperforming OpenAI’s Operator, ChatGPT Atlas, and Gemini 2.5 Computer Use.

• Amazon announced plans to eliminate 14,000 corporate roles to boost efficiency, a move linked to gains from AI and robotics, according to CEO Andy Jassy.

• xAI launched Grokipedia, an AI-powered encyclopedia featuring over 800,000 Grok-generated articles, allowing users to suggest corrections with real-time AI updates.

• GitHub rolled out Agent HQ, a platform that brings together coding agents from Anthropic, OpenAI, Google, Cognition, and xAI, accessible through a unified dashboard.

• OpenAI CEO Sam Altman stated the company expects to reach an “intern-level research assistant” next year and a fully automated AI researcher by 2028.

• Google introduced Pomelli, an experimental Labs project that builds AI-generated marketing campaigns and content directly from a brand’s website.

• Grammarly rebranded as Superhuman and launched ‘Superhuman Go,’ an AI agent suite that automates tasks such as scheduling and research across multiple applications.

• TikTok unveiled Smart Split and AI Outline, tools that use AI to turn longer videos into short clips and create organized content outlines.

• IBM introduced Granite 4.0 Nano, a series of lightweight language models ranging from 350 million to 1.5 billion parameters, optimized for on-device performance.

• Google expanded NotebookLM with features like a larger context window, memory capabilities, customizable chat personas, and higher-quality responses.

• OpenAI released gpt-oss-safeguard, two open models allowing developers to define moderation rules and view the AI’s reasoning when identifying harmful content.

Closing Thoughts

That’s it for us this week.