OSASI Briefs
Ideas, sources and perspectives from the Open Source Appraisal Software Initiative.
Exploring privacy, professional independence and shared ownership of the technology we use.
Each brief is a blend of human and machine. A member reads or listens, sends the moments that mattered, and a draft is assembled from the full source, reviewed, and approved by a person before it is published.
Get new briefs by email
One short email when a new brief is published. Nothing else.
You will get a confirm link first. The list holds only your address; it is never shared or used for anything else, and every email has a one-click unsubscribe.
Prefer a feed reader, Slack or Teams? Follow by RSS.
Before you trust an agent, know what it remembers
A personal AI agent published private work for anyone who knew where to look. Agents can work alongside the analyst, but trusting one means knowing what it shares, what it remembers and what it forgets. Anyone building a system on agents alone has four kinds of memory to master.
The public deserves the backstory
Insiders keep sounding alarms about AI safety, and the people building AI expect far more from it than the public does. The public deserves to know how AI is being used, and the backstory of how it is made. Appraisers have a backstory that deserves greater discovery.
AI versus AI on the web: small firms need a shared defence
AI helped find a hole in the software behind about 70% of the web. The scramble to patch it shows what AI versus AI looks like. Large institutions defend themselves with scale: security teams, and shared knowledge of what actually works. A solo or small appraisal firm can pay for premium hosting and security and keep adding layers, but it still has to work out best practice on its own. Even the appraisal software it relies on may offer less cover than it seems: in a crowded, competitive market, some providers that appear put together may be running lean, and appraisers have no real way to know whether security is getting the attention it needs. A transparent organization built to support the individual can pool that knowledge and those resources, and give each member more protection for less.
Prompt engineering is dead. Domain expertise and deterministic models are not
Imran Ahmad, data scientist and author of 50 Algorithms Every Programmer Should Know, says prompt engineering is dead. His bigger point is that algorithms are not. When a bank declines a mortgage or someone prices a house, the answer has to be reliable and explainable, and an LLM is probabilistic and not a deterministic system.
Cheaper isn't private: the labs solved cost, not sovereignty
The labs have solved cost, The AI Daily Brief says, but not sovereignty. Two answers are on offer: OpenAI's Private Intelligence, and open-source agents like Hermes.
Paying twice: why companies want AI they can own
When the CEO of Palantir says customers want to know they "own the means of production", and Microsoft's CEO warns that companies pay for AI twice, the case for open-weight models is no longer only a startup argument. The show's message is that open models are now an enterprise question about costs and data sovereignty, and that an American open-weight wave led by Reflection AI is about to test it. Both of the loudest voices also have something to sell.
Many models, some open: whose future is this survey describing?
The show's message is that the businesses getting real returns from AI now run a management layer over it: several models, rules for which one does what, a sovereignty strategy, and an eye on what it all costs. Average planned AI investment is up again, to US$210 million. The host tells solo operators why that matters to them too, but never puts a number on it. One detail helps: no organization in the survey earns less than US$50 million a year.
From chicken nuggets to golden nuggets: turn your hard-earned intellect into your own gold mine
The show's message is that building something valuable with AI does not automatically make you valuable. When the employer owns what you build, and the skill to build it is one courses promise to teach in a couple of hours, the value sits in the tool. Lauren Pearl calls it the AI chicken nugget problem. OSASI asks the next question: if you are putting extraordinary effort into building something for someone else, what is your strategy for benefiting? Her closing answer is the golden nugget: relationships and collaboration.
Open weights, two doors: rented through frontier labs, or run on your own hardware
OpenAI Dev Day saw a flurry of new announcements and products, most follow the trends: cheaper good-enough models, always-on agents, and AI work shared across teams. One announcement The AI Daily Brief rates bigger than people are giving it credit for: enterprises can now buy open-weight model inference from Baseten through OpenAI, using their OpenAI spend. That is one door to open weights. A podcast on local AI shows the other: the same kind of model running on hardware you control.
Is OpenAI testing the ceiling on monthly fees?
At its DevDay on September 29, OpenAI reopened its $200 Pro plan to new subscribers with a recalculated allowance and launched a new $500 Pro plan. OpenAI's own explanation is that the allowance is counted in API dollars and its API prices have just been halved, so subscribers still get more work done. The other question: is this also a test of how high a monthly fee will go?
Neither boom nor doom: the AI moderates
New polling puts Americans among the most worried people in the world about AI and among the least excited. The AI Daily Brief argues that most people are still in the middle, and that a group it calls the AI moderates is starting to find its voice. OSASI recognises the position: responsible adoption, not reaction, is the way forward.
New builders, stable foundations
In consecutive weeks in July, Value Driven Data Science gave data professionals two pieces of advice. First, stop reading about AI and build something with it, because the leap is smaller than it looks. Second, do not build your future entirely on rented models, because today's prices are subsidised. Together they describe a shift in who gets to build: those with new insight and new methods, and the recipe to put the best of innovation on a new but stable foundation on an even keel.
It's fall. How do you budget AI for 2027?
In June, The Artificial Intelligence Show asked whether the era of affordable AI is over. Its answer: nobody knows how to price this yet, heavy users are being carried by light ones, and falling costs may rescue the flat-price plan. The host ended on a question for the fall: how do you even begin to budget AI for 2027? The fall has arrived.
$600 to $7,000 per day: who is paying for AI?
The show's message is that cost is the story: OpenAI's new models halve API prices, and every drop in price opens up more work for AI. OpenAI's own figure shows where that leads: valued at API prices, its median researcher uses more than $600 of tokens a day, and $7,000 at the 90th percentile.
Below the waterline: who is measuring the freeboard?
Choosing which AI tools to buy is the easy part of an AI strategy. The hard part, the show says, is everything beneath the tools, from platform and inference to hardware and sovereignty. Follow the layers down and every one of them ends at the same question for a profession entrusted with confidential files: who inspects for contaminated ballast water discharge?
Hybrid agentic AI powerhouse, or human-AI slop soup?
Agentic AI is sold as doing the job faster. A consultant who builds hybrid organisations makes the better case: quality must remain the key focus, and not the inverse. Humans and AI working together in ways that neither could manage alone.
Is better the enemy of good enough?
The frontier race wants ever more capable models. Appraisal work needs models that are accountable, auditable and fit for the task. Two segments from one morning show sharpen the difference.
Useful AI needs accountable boundaries.
Privacy depends on more than where information is stored. It also depends on what an AI system can access, what it can do with that information, and who remains in control.
Our data. Our knowledge. Our control.
The case for shared AI infrastructure starts with privacy: keeping confidential information and professional knowledge under the control of those entrusted with it.