Sapient Artifice Newsletter 8
AI Trends in 2025
The short version: This past month is a continuation of the AI narrative - from the lab to the legislature, from deep space to deepfakes, the frontier is expanding with a velocity that leaves governance, ethics, and public oversight behind. The last month of AI rollout is the progression of a year-long pattern of unchecked acceleration.
We began 2025 watching AI reshape protein engineering, medical diagnostics, and national security. By mid-year, we flagged red-teaming fragility, cultural backlash, and uneven federal policy. Now, as we approach season’s close, the feedback loop has become clear: AI capability accelerates, power consolidates, oversight decays, and trust erodes.
And yet, if you only read the headlines, you might miss something crucial - the system isn’t broken by accident. It’s broken by design.
News highlights
If you'd have only read the news at the beginning of this past month, you would think things are starting to cool off. From MIT to the NYT the word was out that AI isn't making money fast enough. That entry level work is going away, a bubble is forming, and the economy may be in trouble. However, as we conclude this month, the forces that power AI are still holding their ground at all time highs.
With this continued acceleration the AI energy footprint is on our collective minds. Google came out with some per prompt cost estimates and discussion regarding their gains in efficiency this year. The efficiency increases we have seen like this over the last few months should remind us that direct extrapolations of our long term energy footprint may actually be somewhat misleading right now. Even more, over these last months the number of breakthrough R&D projects with the potential to significantly reduce system requirements continue to pile up, like this optical generative model and this low power microwave based chip, as this months examples. We have seen time and time again, technology has a way of solving technical problems - that's rarely our bottleneck.
What seems more immediately pressing is the rapidly increasing roles, responsibilities, and authority being granted to computer systems. Microsoft spoke up this month offering some common sense - first, anchoring society in reality by taking the important position regarding AI non-sentience and non-personhood - a position we have discussed the importance of many times before, and then secondly, by releasing this study that takes a serious look at how AI will impact job migration, which is a subject we were just recently discussing.
We close our news highlights with some of the most amazing developments in AI this month to remind ourselves how amazing and powerful it can be when being used for our collective social good. This month AI was successfully collaborating in the design of experiments in quantum optics - helping to push our understanding of fundamental physics. We saw a highschooler work with AI to discover 1.5 million lost space objects as part of his summer research program. And we met our first AI powered stethoscope that increases the ability to recognize heart failure by 2.3 times and to detect abnormalities by 3.5 times.
Legislation & policy
At the Federal level Senator Cruz filed the SANDBOX Act. For many, this looks like a new pay-to-play gamification of AI regulation. The Act does mandate public disclosure of processes and requires detailed applications including risk mitigation plans. The larger concern however remains we haven't even established a federally coherent AI regulatory framework yet, and we are already legislating work-arounds for those with the power to push?
As we prepare to move into the fall legislative session, there are a few places already this year where the states have made a loud and unified statement (1) We don't want to be deepfaked. Fourty-eight of the fifty US States now have deep fake laws passed. (2) We are not ok with our children being harmed. Attorney General Skrmetti Leads 44 States in Demanding Companies End Predatory AI Interactions with Kids. In all ways that the states can continue to unify there efforts, the tighter our tapestry will be.
The U.S. is not failing to lead AI policy. It’s choosing not to. And the world is choosing paths without us.
Company update
Here around the lab we have been working feverishly all month! It was time to assess where we stood with our efforts this year - so our wave two user survey was sent out. Thanks to everyone who answered! We are happy to say that the sentiment has been positive with 4.6/5 overall satisfaction, and 9.4/10 would recommend to friends and colleagues.
During this time, we also completed the first generation of our commercial gateway. We are excited to say we will now begin opening it up together with our desktop application, mage lab, into public beta.
Beginning today the desktop app for Mac (Apple Silicon) · Windows · Linux (deb) & Oauth based gateway accounts are now available. In the coming days, we will be adding broader account creation methods and additional models to the gateway. Stay tuned!
As we continue rolling out our software and services, we will write more about it here, as well as opening up our Discord. We look forward to talking to you there and seeing what you do with mage lab. We hope you will love it!
Behind every system failure is a design failure. AI is not just breaking - it’s breaking the wrong way. And the deeper we go down that rabbit hole, the more foundational the fragility becomes. We do not claim to have all the answers, but we promise we are building with these most important questions, concerns, and values driving our choices. And that with the support of our community, we are dedicated to actively fighting the good fight in every way we can. This includes developing good ole fashion goods and services that do not aim to productize you - instead they will empower you in the coming age to remain free and in control of your own path. More to come.
Sapient Artifice