Kia ora {{ First name | friend }},
We’re exploring the light and shadow of technology, with a focus on efficient AI workflows to free up your time.
This week’s edition features:
USA govt and AI companies unite
Human-centric AI lab
Tiny swimming AI robots
Google launches new products
ChatGPT Image updates
Announcement: Starting January, this email will now be once every two weeks, instead of every week. It’ll also come on 7pm Mondays, not Sundays. This is to fit my lifestyle better, and ensure each newsletter is filled with quality, not slop.
Thanks for being here.
Feature: USA Heavy-Hitters Team Up with Govt for AI Alliance
Overview
The U.S. Department of Energy signed collaboration agreements with 24 organizations to advance the Genesis Mission, a national effort to embed artificial intelligence into core scientific research. Key partners include OpenAI, Google, Anthropic, Intel, Microsoft, xAI, Palantir, and Nvidia. The goal is speed. Faster discovery, faster experimentation, and faster deployment across energy, materials science, and national security.
What they’re doing
Genesis focuses on using AI systems to automate experiment design, run large scale simulations, and build predictive models across DOE national laboratories. This shifts science from manual trial and error toward continuous, machine assisted discovery. Researchers spend less time setting up experiments and more time interpreting results.
These collaborations operate through non binding agreements. No procurement bottlenecks. No slow contracting cycles. Teams share expertise, infrastructure access, and research workflows. This structure lets the DOE move at AI speed while keeping work aligned with public interest goals.
Why they’re doing it
The USA and China are supposedly neck-and-neck in a race for global (AI) dominance. Genesis also ties into broader US policy. The initiative supports national AI leadership, supply chain resilience, and scientific competitiveness. Energy systems, advanced materials, and nuclear security all benefit from the outcomes of these partnerships. AI now sits at the center of US science strategy, not on the edge.
The human layer
I have some reservations around this. All of it feels strangely familiar to the beginning of a sci-fi plot… “Let’s just put all the most powerful and well-funded tech companies alongside a gung-ho US govt, in the name of science… What’s the worst that could happen?”
Well, who knows actually what the worst that could happen is. However, it’s not hard to imagine that some of this could be on the radar…
Increase in biotech development velocity
Expansion of digital survelliance
Increased reliance on digital services for every day life
Consolidation of scientific power among a small group of vendors
Decrease in public opinion shaping the future of tech
The issue I feel lies not in the takeover itself, but the consciousness behind it. I’ve recently heard the digital corporate infiltration of our lives as a form of colonisation 2.0. But now we’re all being colonised, all at once, by unelected and unregulated leaders, who seek to implement things that perhaps not all affected people agree with?
Also, why isn’t Meta on the list? Poor Zuckerberg… although all his mates from Time' Magazine’s People of the Year are on the list.
In Lighter News: Human-Centric AI Lab
Figure AI’s founder and CEO Brett Adcock is launching a separate artificial intelligence lab called Hark, backed by $100 million of his personal capital. The new lab will pursue “human-centric AI” that reasons proactively and improves continuously, with its own GPU compute cluster now live. Adcock continues to lead Figure while building Hark as a distinct research effort aimed at creating systems that think deeper than typical large language models.
The move reflects a broader trend in the AI world where established founders are spinning out focused research initiatives outside their core companies. Hark’s mission is to explore architectures and paradigms that executives believe mainstream labs overlook. His language suggests ambitions toward autonomous reasoning, embodied intelligence, and AI that adapts over time with minimal supervision.
Figure AI itself already stands at the forefront of the rising humanoid robotics sector. The company has raised over $1 billion and carries a multi-billion dollar valuation. Figure is building general-purpose robots for industrial and service roles, targeting high-volume production and real-world deployments.
This new lab raises strategic questions about how AI innovation is structured. Adcock’s dual leadership of a robotics company and a standalone AI research outfit blurs lines between product-driven engineering and pure research. If Hark yields breakthroughs, the spin-off model could reshape how founders incubate next-generation AI systems outside existing frameworks.
Tiny AI Robots, Smaller Than a Grain of Salt… Hmm
Researchers at the University of Pennsylvania and University of Michigan have created the world’s smallest fully programmable, autonomous robots.
And they can swim…
“Microscopic swimming machines can independently sense and respond to their surroundings, operate for months and cost just a penny each.”
"Barely visible to the naked eye, each robot measures about 200 by 300 by 50 micrometers, smaller than a grain of salt. Operating at the scale of many biological microorganisms, the robots could advance medicine by monitoring the health of individual cells and manufacturing by helping construct microscale devices."
"Powered by light, the robots carry microscopic computers and can be programmed to move in complex patterns, sense local temperatures and adjust their paths accordingly."
It’s a funny thing to think that even right now, we could have one of these little buggers inside us, and not even know…
Here’s a few interesting technology updates:
Bill Gates makes yet another feature for his digital override
OpenAI launched GPT-5.2-Codex, which comes with updated coding capabilities and strengthened cybersecurity. Read more.
Mistral launched an OCR platform which enables you to take a picture and have it “read” by AI with incredible accuracy. This tech has existed for a long time but is notoriously slow, perhaps this could be the one to change it? Check it out.
Elon Musk targets artificial general intelligence by 2026, he told staff during an internal all hands meeting at xAI that the company targets artificial general intelligence on an aggressive timeline, with internal expectations pointing to next year. He framed the outcome as a survival test, and if they can make it through the next few years, they’ll come out on top. Read more.
Google’s vibe-coding tool Opal comes to Gemini, and is now integrated into the Gemini web app. Continue reading.
Amazon reportedly in talks to invest $10B in OpenAI, which would value the company at over $500 billion. The circular deal pattern continues. Check it out.
“Slop Evader” is a browser extension that is gaining attention because it lets people browse the web as if it were still 2022. It blocks anything created after November 30, 2022, the day ChatGPT was released, and only shows pages from before generative AI reshaped the internet. Install it here.
Google launched a series Gemini-powered translation upgrades, including new features that enable you to translate live speech in any headphones. Basically Google translate in your ears. Read more.
ChatGPT Images just got better with GPT Image 1.5
OpenAI released GPT Image 1.5, a major rebuild of image generation inside ChatGPT. The release targets precision, speed, and repeatable creative work, not novelty images. This follows strong traction from Google’s recent image tooling and signals a shift toward production grade visual systems.
Finally, we can now adjust individual elements within an image, unlike previous models. GPT Image 1.5 modifies only the element you request while preserving lighting, composition, and facial identity. This removes the need to regenerate full images to fix small details, a long standing frustration in AI design.
OpenAI also introduced a dedicated Images workspace. The new tab functions as a lightweight creative studio with preset styles, prompt templates, and inspiration flows. Many transformations work without typed prompts, lowering friction for non technical users. I turned a picture of my parents’ dog, Louie, into a keychain, using one of these.
Additionally, image generation runs roughly four times faster, enabling near real time iteration. This supports use cases like design review, marketing drafts, product mockups, and content testing without breaking flow.
Developers get access through the GPT Image 1.5 API. The model works across ChatGPT without manual selection, simplifying integration into apps and pipelines.
On the competitive front, Google continues pushing visual tooling through Gemini, including Opal for vibe based creation. Early testers (including myself) still rate Google’s image output higher in some cases.
The gap continues to narrow.
{{ First name | friend }}, thanks for dropping in again.
Forward this to a friend if you found it useful.
Stay human,
Billy