<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Frontier Leaf</title><link>https://frontierleaf.com/</link><description>A field journal of new AI developments.</description><language>en</language><atom:link href="https://frontierleaf.com/feed.xml" rel="self" type="application/rss+xml"/><atom:link rel="hub" href="https://pubsubhubbub.appspot.com/"/><item><title>AI models resolve decades-old math problems, shifting student focus to regulation</title><link>https://frontierleaf.com/section/models/ai-models-resolve-decades-old-math-problems/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/ai-models-resolve-decades-old-math-problems/</guid><description>OpenAI’s recent success with complex mathematics suggests students must pivot from traditional study to shaping AI policy and educational reform.</description><pubDate>Sun, 01 Nov 2026 19:08:02 +0000</pubDate><category>Models</category></item><item><title>Anthropic filing reveals R&amp;D consumes majority of compute spend</title><link>https://frontierleaf.com/section/models/anthropic-filing-reveals-r-d-consumes-majority-of/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/anthropic-filing-reveals-r-d-consumes-majority-of/</guid><description>IPO documents show training and research absorb most of the $8.9bn bill, while customer serving costs remain low.</description><pubDate>Sun, 01 Nov 2026 17:20:18 +0000</pubDate><category>Models</category></item><item><title>Google scales AI Mode ads globally as monetisation concerns weigh on stock</title><link>https://frontierleaf.com/section/models/google-scales-ai-mode-ads-globally-as/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/google-scales-ai-mode-ads-globally-as/</guid><description>Alphabet expands advertising in its Gemini-powered search feature across ten new markets while acknowledging lower revenue per query compared to traditional search.</description><pubDate>Sun, 01 Nov 2026 16:43:54 +0000</pubDate><category>Models</category></item><item><title>Melodysheep creates cinematic exploration with AI titled Binary Dreams</title><link>https://frontierleaf.com/section/models/melodysheep-creates-cinematic-exploration-with-ai/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/melodysheep-creates-cinematic-exploration-with-ai/</guid><description>A Bing News report describes Binary Dreams, a cinematic work created by Melodysheep and AI, where an advanced AI explores its origins and future.</description><pubDate>Sun, 01 Nov 2026 10:10:20 +0000</pubDate><category>Models</category></item><item><title>OpenAI CEO argues home PC RAM is a poor use of scarce memory supply</title><link>https://frontierleaf.com/section/models/openai-ceo-argues-home-pc-ram-is-a-poor-use-of/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/openai-ceo-argues-home-pc-ram-is-a-poor-use-of/</guid><description>Sam Altman claims local inference wastes limited DRAM that should serve cloud data centres at higher utilisation rates.</description><pubDate>Sun, 01 Nov 2026 07:44:16 +0000</pubDate><category>Models</category></item><item><title>Netgear RAX5 reaches end of service after final security patch for local network flaws</title><link>https://frontierleaf.com/section/models/netgear-rax5-reaches-end-of-service-after-final/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/netgear-rax5-reaches-end-of-service-after-final/</guid><description>Patches are available for specific Netgear models affected by two medium severity vulnerabilities that permit local network attackers to run commands or alter settings.</description><pubDate>Sun, 01 Nov 2026 06:37:31 +0000</pubDate><category>Models</category></item><item><title>Search fully restored in ChatGPT Work, Dots and thinking experiences following OpenAI outage</title><link>https://frontierleaf.com/section/models/search-fully-restored-in-chatgpt-work-dots-and/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/search-fully-restored-in-chatgpt-work-dots-and/</guid><description>OpenAI has resolved the search disruptions affecting ChatGPT Work, Dots and certain thinking experiences, with all services now operating normally.</description><pubDate>Sun, 01 Nov 2026 06:14:53 +0000</pubDate><category>Models</category></item><item><title>Intuit launches AI-focused accounting curriculum to bridge skills gap</title><link>https://frontierleaf.com/section/models/intuit-launches-ai-focused-accounting-curriculum/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/intuit-launches-ai-focused-accounting-curriculum/</guid><description>Intuit introduces free educational programme pairing AI fluency with advisory training to prepare graduates for client-facing roles.</description><pubDate>Sun, 01 Nov 2026 05:39:54 +0000</pubDate><category>Models</category></item><item><title>Google Workspace enables native Markdown editing in Docs for developers and AI agents</title><link>https://frontierleaf.com/section/agents/google-workspace-enables-native-markdown-editing/</link><guid isPermaLink="true">https://frontierleaf.com/section/agents/google-workspace-enables-native-markdown-editing/</guid><description>Google Workspace now supports native .md file editing in Docs, allowing developers and AI agents to collaborate without manual conversions or loss of Git compatibility.</description><pubDate>Sun, 01 Nov 2026 05:35:38 +0000</pubDate><category>Agents</category></item><item><title>Google A2UI standardises agent interfaces using declarative data structures</title><link>https://frontierleaf.com/section/agents/google-a2ui-standardises-agent-interfaces-using/</link><guid isPermaLink="true">https://frontierleaf.com/section/agents/google-a2ui-standardises-agent-interfaces-using/</guid><description>Google&#x27;s new open standard enables agents to render rich UIs across frameworks by describing interfaces as data rather than executable code.</description><pubDate>Sun, 01 Nov 2026 05:32:22 +0000</pubDate><category>Agents</category></item><item><title>Google ADK enables non-blocking voice agents with gemini-3.8-live</title><link>https://frontierleaf.com/section/agents/google-adk-enables-non-blocking-voice-agents-with/</link><guid isPermaLink="true">https://frontierleaf.com/section/agents/google-adk-enables-non-blocking-voice-agents-with/</guid><description>Developers can eliminate dead air in voice conversations by using streaming tools and response scheduling in the open-source Agent Development Kit.</description><pubDate>Sun, 01 Nov 2026 05:06:17 +0000</pubDate><category>Agents</category></item><item><title>OpenAI: Fair Use Law Powers AI Progress &amp; Helps Disabled People Access Content</title><link>https://frontierleaf.com/section/models/openai-fair-use-law-powers-ai-progress-helps/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/openai-fair-use-law-powers-ai-progress-helps/</guid><description>OpenAI shares excerpts from legal filings arguing that fair use supports AI innovation and helps disabled users access content.</description><pubDate>Sun, 01 Nov 2026 04:25:55 +0000</pubDate><category>Models</category></item><item><title>OpenAI Patches ChatGPT Flaw Letting Code Steal Data, Awards $5k Bounty</title><link>https://frontierleaf.com/section/models/openai-patches-chatgpt-flaw-letting-code-steal/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/openai-patches-chatgpt-flaw-letting-code-steal/</guid><description>Parameter Labs detailed how untrusted code in ChatGPT previews could escape sandboxing and access private user data through a complex exploit chain.</description><pubDate>Sun, 01 Nov 2026 04:09:19 +0000</pubDate><category>Models</category></item><item><title>Dartmouth provost resigns over undisclosed AI use in academic writing</title><link>https://frontierleaf.com/section/models/dartmouth-provost-resigns-over-undisclosed-ai-use/</link><guid isPermaLink="true">https://frontierleaf.com/section/models/dartmouth-provost-resigns-over-undisclosed-ai-use/</guid><description>Santiago Schnell stepped down after an inquiry found he failed to reveal his use of large language models in a journal paper and an opinion piece.</description><pubDate>Sat, 31 Oct 2026 20:07:49 +0000</pubDate><category>Models</category></item></channel></rss>