<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"><channel><title>The Raw Logs — #ai</title><link>https://therawlogs.com/tag/ai</link><description>Top takes tagged #ai.</description><language>en</language><lastBuildDate>Tue, 08 Sep 2026 16:37:47 GMT</lastBuildDate>    <item><title>#ai — Samsung Opens CONNECT at IFA 2026 for On-Device Contextual AI Living.</title><link>https://therawlogs.com/take/15-5</link><guid>https://therawlogs.com/take/15-5</guid><pubDate>Wed, 02 Sep 2026 17:00:13 GMT</pubDate><description><![CDATA[Samsung Opens CONNECT at IFA 2026 for On-Device Contextual AI Living.: Samsung opened its IFA 2026 showcase with on-device AI across appliances. Streaming home video and voice feeds to cloud servers leaks private data and wastes bandwidth. Local chips process household routines privately. [#ai]]]></description></item>    <item><title>#ai — MIT neural physics model simulates disasters never seen in training</title><link>https://therawlogs.com/take/13-4</link><guid>https://therawlogs.com/take/13-4</guid><pubDate>Sat, 29 Aug 2026 16:25:06 GMT</pubDate><description><![CDATA[MIT neural physics model simulates disasters never seen in training: MIT built a neural model that predicts structural material failure in extreme disasters it never saw in training data. Standard models fail on rare edge cases. Enforcing physical stress rules directly in model architecture accurately simulates unobserved physical catastrophes. [#ai]]]></description></item>    <item><title>#ai — Cisco introduces hardware-enforced network enclaves for sovereign AI defense</title><link>https://therawlogs.com/take/10-2</link><guid>https://therawlogs.com/take/10-2</guid><pubDate>Wed, 26 Aug 2026 16:15:43 GMT</pubDate><description><![CDATA[Cisco introduces hardware-enforced network enclaves for sovereign AI defense: Cisco launched Cisco IQ to isolate AI model traffic directly at network hardware switches. Software-only guardrails fail against prompt injection and tool spoofing. Hardware-level network enclaves physically block data exfiltration, keeping internal code and data contained. [#ai]]]></description></item>    <item><title>#ai — Checking behavior during execution prevents AI agents from failing.</title><link>https://therawlogs.com/take/9-4</link><guid>https://therawlogs.com/take/9-4</guid><pubDate>Tue, 25 Aug 2026 16:40:45 GMT</pubDate><description><![CDATA[Checking behavior during execution prevents AI agents from failing.: Alice raised $140M for in-line behavioral filters that inspect agent tool parameters before execution. Post-event logs only alert you after damage occurs. Real-time semantic interceptors evaluate intent in milliseconds and block unapproved database writes. [#ai]]]></description></item>    <item><title>#ai — AI costs move from training to real-time inference.</title><link>https://therawlogs.com/take/6-3</link><guid>https://therawlogs.com/take/6-3</guid><pubDate>Sat, 22 Aug 2026 16:49:15 GMT</pubDate><description><![CDATA[AI costs move from training to real-time inference.: Data shows over 75% of AI spend now goes to using models, not training. When queries cost millions monthly, training efficiency matters less. Success depends on query speed and cost. Shift budget from big training jobs to inference hardware and software to cut daily costs. [#ai]]]></description></item>    <item><title>#ai — Training AI agents in continuous virtual physics and multi‑agent economies.</title><link>https://therawlogs.com/take/5-4</link><guid>https://therawlogs.com/take/5-4</guid><pubDate>Fri, 21 Aug 2026 16:29:25 GMT</pubDate><description><![CDATA[Training AI agents in continuous virtual physics and multi‑agent economies.: DeepMind trains models with ongoing simulations where many agents interact in a virtual economy and 3D physics. Testing agents only on static text makes code fail in real tasks. Multi‑agent settings need tools for resource handling, planning and teamwork over many steps. [#ai]]]></description></item></channel></rss>