<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel>
<title>OSASI Briefs</title><link>https://briefs.osasi.org/</link><atom:link href="https://briefs.osasi.org/feed.xml" rel="self" type="application/rss+xml"/>
<description>Ideas, sources and perspectives from the Open Source Appraisal Software Initiative.</description><language>en-ca</language>
<lastBuildDate>Wed, 28 Oct 2026 19:00:00 GMT</lastBuildDate>
<item><title>Before you trust an agent, know what it remembers</title><link>https://briefs.osasi.org/posts/2026-10-28-know-what-your-agent-remembers.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-10-28-know-what-your-agent-remembers.html</guid><pubDate>Wed, 28 Oct 2026 19:00:00 GMT</pubDate><description>Meta&apos;s Muse agent left shared chats findable on the web. The AI Show asks whether today&apos;s skills will outlive the next model. Value Driven Data Science lays out the memory an agent needs. Where do agents fit in an appraiser&apos;s work? OSASI Commentary.</description></item>
<item><title>The public deserves the backstory</title><link>https://briefs.osasi.org/posts/2026-10-24-the-public-deserves-the-backstory.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-10-24-the-public-deserves-the-backstory.html</guid><pubDate>Sat, 24 Oct 2026 19:00:00 GMT</pubDate><description>A former OpenAI safety lead says the industry is not being safe enough. Anthropic finds its own staff and the public far apart on where AI is heading. Karen Hao tells the darker side. What do appraisers owe the people who rely on their work? OSASI Commentary.</description></item>
<item><title>AI versus AI on the web: small firms need a shared defence</title><link>https://briefs.osasi.org/posts/2026-10-23-ai-vs-ai-shared-defence.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-10-23-ai-vs-ai-shared-defence.html</guid><pubDate>Fri, 23 Oct 2026 19:00:00 GMT</pubDate><description>AI helped find PHP2Shell, a hole in the software behind about 70% of websites. The AI Daily Brief calls it the start of an AI versus AI era. Big institutions can buy scale to defend themselves; what about solo and small appraisal firms? OSASI Commentary.</description></item>
<item><title>Prompt engineering is dead. Domain expertise and deterministic models are not</title><link>https://briefs.osasi.org/posts/2026-10-20-domain-expertise-deterministic.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-10-20-domain-expertise-deterministic.html</guid><pubDate>Tue, 20 Oct 2026 19:00:00 GMT</pubDate><description>On Value Driven Data Science, Imran Ahmad says prompt engineering is dead but algorithms still matter: a mortgage decision or a house price needs a model that shows its work. A CBC test that put one house through six appraisals shows why consistency is the real issue. OSASI Commentary.</description></item>
<item><title>Cheaper isn&apos;t private: the labs solved cost, not sovereignty</title><link>https://briefs.osasi.org/posts/2026-10-13-cheaper-isnt-private.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-10-13-cheaper-isnt-private.html</guid><pubDate>Tue, 13 Oct 2026 19:00:00 GMT</pubDate><description>Claude Haiku 5.5 makes the cheapest capable models American. OpenAI&apos;s Private Intelligence and the open-source Hermes Agent offer two answers on control. What does each keep private? OSASI Commentary.</description></item>
<item><title>Paying twice: why companies want AI they can own</title><link>https://briefs.osasi.org/posts/2026-10-06-ai-they-can-own.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-10-06-ai-they-can-own.html</guid><pubDate>Tue, 06 Oct 2026 19:00:00 GMT</pubDate><description>The AI Daily Brief reads the Axios scoop on Reflection AI&apos;s coming American open-weight model and hears enterprise demand for AI they can control: Palantir&apos;s Alex Karp on owning the means of production, Satya Nadella on paying for intelligence twice. Is your data safe with the frontier labs, and does a US server make a Chinese model safe? OSASI Commentary.</description></item>
<item><title>Many models, some open: whose future is this survey describing?</title><link>https://briefs.osasi.org/posts/2026-10-04-many-models-whose-future.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-10-04-many-models-whose-future.html</guid><pubDate>Sun, 04 Oct 2026 19:00:00 GMT</pubDate><description>The AI Daily Brief reads KPMG&apos;s Q3 Global AI Pulse: the leaders are routing work across several models, writing sovereignty strategies and planning US$210 million each for AI next year. The host gives solo operators two reasons to care, but every organization in the survey earns at least US$50 million a year. OSASI Commentary.</description></item>
<item><title>From chicken nuggets to golden nuggets: turn your hard-earned intellect into your own gold mine</title><link>https://briefs.osasi.org/posts/2026-10-01-chicken-nuggets-golden-nuggets.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-10-01-chicken-nuggets-golden-nuggets.html</guid><pubDate>Thu, 01 Oct 2026 19:00:00 GMT</pubDate><description>Value Driven Data Science with Lauren Pearl: the person who invents the chicken nugget is rarely the one who gets rich from it, and building with AI is now easy enough that the tool, not its builder, can take the value. Safety, judgement, rare skills and, above all, collaboration are where it still lives. OSASI Commentary.</description></item>
<item><title>Open weights, two doors: rented through frontier labs, or run on your own hardware</title><link>https://briefs.osasi.org/posts/2026-09-30-open-weights-two-doors.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-30-open-weights-two-doors.html</guid><pubDate>Wed, 30 Sep 2026 19:00:00 GMT</pubDate><description>At DevDay OpenAI opened a marketplace where enterprises can buy open-weight model inference from Baseten with their OpenAI spend. A podcast on local AI shows the other door: open models on hardware you control, with the cloud kept for the hard thinking. OSASI Commentary.</description></item>
<item><title>Is OpenAI testing the ceiling on monthly fees?</title><link>https://briefs.osasi.org/posts/2026-09-29-testing-the-ceiling.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-29-testing-the-ceiling.html</guid><pubDate>Tue, 29 Sep 2026 19:00:00 GMT</pubDate><description>At DevDay OpenAI reopened its $200 Pro plan with half the allowance counted in API dollars and launched a $500 tier. OpenAI&apos;s reason is that API prices have halved. OSASI asks whether this is also a test of how high a monthly fee the market will carry. OSASI Commentary.</description></item>
<item><title>Neither boom nor doom: the AI moderates</title><link>https://briefs.osasi.org/posts/2026-09-28-the-ai-moderates.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-28-the-ai-moderates.html</guid><pubDate>Mon, 28 Sep 2026 19:00:00 GMT</pubDate><description>Americans are among the world&apos;s most worried about AI, and among the least excited. The AI Daily Brief finds a quieter group forming between the accelerationists and the doomers: Francis Fukuyama, Kapoor and Narayanan, and Jeffrey Katzenberg. OSASI Commentary.</description></item>
<item><title>New builders, stable foundations</title><link>https://briefs.osasi.org/posts/2026-09-27-new-builders-stable-foundations.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-27-new-builders-stable-foundations.html</guid><pubDate>Sun, 27 Sep 2026 19:00:00 GMT</pubDate><description>Two Value Driven Data Science episodes with Nicholas Kelly: data professionals can now build their own software with AI, and should own at least part of the machine it runs on before subsidised prices end. The shift favours those with new insight and new methods, who build the best of innovation on a stable foundation, on an even keel. OSASI Commentary.</description></item>
<item><title>It&apos;s fall. How do you budget AI for 2027?</title><link>https://briefs.osasi.org/posts/2026-09-24-budgeting-ai-for-2027.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-24-budgeting-ai-for-2027.html</guid><pubDate>Thu, 24 Sep 2026 19:00:00 GMT</pubDate><description>In June, The Artificial Intelligence Show asked whether the era of affordable AI is over: flat-price plans that heavy users push far past their price, paid for by light users. Three months on, the price cuts it anticipated have arrived, and budgeting season with them. OSASI Commentary.</description></item>
<item><title>$600 to $7,000 per day: who is paying for AI?</title><link>https://briefs.osasi.org/posts/2026-09-24-who-pays-for-the-tokens.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-24-who-pays-for-the-tokens.html</guid><pubDate>Thu, 24 Sep 2026 19:00:00 GMT</pubDate><description>OpenAI says its median researcher uses $600 of tokens a day, and $7,000 at the 90th percentile, valued at API prices that were just cut in half. A race to dominance, and why OSASI favours reproducible code tools that rely less on generative AI, for a more sustainable system. OSASI Commentary.</description></item>
<item><title>Below the waterline: who is measuring the freeboard?</title><link>https://briefs.osasi.org/posts/2026-09-23-below-the-waterline.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-23-below-the-waterline.html</guid><pubDate>Wed, 23 Sep 2026 19:00:00 GMT</pubDate><description>Choosing which AI tools to buy is the easy part of an AI strategy; the hard part is everything beneath them. Victor Coimbra&apos;s four layers end at governance, security, privacy and sovereignty, and at who is measuring the freeboard. OSASI Commentary.</description></item>
<item><title>Hybrid agentic AI powerhouse, or human-AI slop soup?</title><link>https://briefs.osasi.org/posts/2026-09-22-hybrid-agentic-powerhouse-or-slop-soup.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-22-hybrid-agentic-powerhouse-or-slop-soup.html</guid><pubDate>Tue, 22 Sep 2026 19:00:00 GMT</pubDate><description>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. OSASI Commentary.</description></item>
<item><title>Is better the enemy of good enough?</title><link>https://briefs.osasi.org/posts/2026-09-21-better-the-enemy-of-good-enough.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-21-better-the-enemy-of-good-enough.html</guid><pubDate>Mon, 21 Sep 2026 19:00:00 GMT</pubDate><description>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. OSASI Commentary.</description></item>
<item><title>Useful AI needs accountable boundaries.</title><link>https://briefs.osasi.org/posts/2026-09-20-useful-ai-needs-accountable-boundaries.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-20-useful-ai-needs-accountable-boundaries.html</guid><pubDate>Sun, 20 Sep 2026 19:00:00 GMT</pubDate><description>Privacy depends on more than where information is stored: what an AI system can access, what it can do with it, and who remains in control. OSASI Commentary.</description></item>
<item><title>Our data. Our knowledge. Our control.</title><link>https://briefs.osasi.org/posts/2026-09-19-owned-models-and-appraisal.html</link><guid isPermaLink="true">https://briefs.osasi.org/posts/2026-09-19-owned-models-and-appraisal.html</guid><pubDate>Sat, 19 Sep 2026 19:00:00 GMT</pubDate><description>Shared AI infrastructure for privacy, security and control over confidential information and professional knowledge. OSASI Commentary.</description></item>
</channel></rss>
