<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/">
<channel>
<title>AI Frontier Post</title>
<link>https://aifrontierpost.com/</link>
<description>Dispatches from the AI frontier: AI news, research explainers, frontier lab coverage, and product comparisons.</description>
<language>en</language>
<lastBuildDate>Wed, 18 Nov 2026 05:49:06 +0000</lastBuildDate>
<atom:link href="https://aifrontierpost.com/feed.xml" rel="self" type="application/rss+xml" xmlns:atom="http://www.w3.org/2005/Atom"/>
<item>
<title>Nvidia doubles revenue to $115.8B in its first report without Jensen Huang — and guides to $130B</title>
<link>https://aifrontierpost.com/articles/nvidia-q3-fy2027-earnings-115-8b-beat/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/nvidia-q3-fy2027-earnings-115-8b-beat/</guid>
<pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Sofia Almeida</dc:creator>
<description>Nvidia’s first earnings report since Jensen Huang’s resignation beat every forecast that mattered: $115.8 billion in revenue, up 103% in a year, and a $130 billion outlook for next quarter. The catch sits in the margins — and in the chair’s flat statement that Huang is not coming back.</description>
<enclosure url="https://aifrontierpost.com/articles/nvidia-q3-fy2027-earnings-115-8b-beat/cover.webp" type="image/webp"/>
</item>
<item>
<title>Fock Ball: the fake sports league Claude built that plays, broadcasts and reports on itself</title>
<link>https://aifrontierpost.com/articles/fock-ball-claude-self-running-sports-league/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/fock-ball-claude-self-running-sports-league/</guid>
<pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Sofia Almeida</dc:creator>
<description>A product designer spent five months with Claude Code building a fictional football league — 30 teams, 38 years of simulated history, a real physics engine — and on Tuesday it went live with a nightly 3D broadcast. The league now runs itself.</description>
<enclosure url="https://aifrontierpost.com/articles/fock-ball-claude-self-running-sports-league/cover.webp" type="image/webp"/>
</item>
<item>
<title>Microsoft’s MAI Code 2 claims 96% of Claude Opus 5.5 at 5% of the price — and developers are pushing back</title>
<link>https://aifrontierpost.com/articles/microsoft-mai-code-2-claude-opus-price-claim/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/microsoft-mai-code-2-claude-opus-price-claim/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Priya Nair</dc:creator>
<description>Microsoft unveiled seven in-house MAI models at Ignite, led by a coding model it says nearly matches Claude Opus 5.5 for a twentieth of the cost. The catch, buried in the fine print and all over Reddit: the price is promotional, the tests are Microsoft’s own, and the developers it needs to convince aren’t convinced yet.</description>
<enclosure url="https://aifrontierpost.com/articles/microsoft-mai-code-2-claude-opus-price-claim/cover.webp" type="image/webp"/>
</item>
<item>
<title>Turn a technical book into a Claude Code skill: hands-on with book-to-skill</title>
<link>https://aifrontierpost.com/articles/book-to-skill-pdf-to-agent-skill-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/book-to-skill-pdf-to-agent-skill-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Priya Nair</dc:creator>
<description>book-to-skill turns any technical book into a structured agent skill — frameworks, decision rules, and anti-patterns your coding agent loads on demand. This tutorial installs it, extracts a real book and a real PDF, and generates a working skill, with every command verified.</description>
<enclosure url="https://aifrontierpost.com/articles/book-to-skill-pdf-to-agent-skill-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>OpenAI releases GPT-6.2 Astra and Sol: cheaper, faster — and its first model rated High for AI self-improvement</title>
<link>https://aifrontierpost.com/articles/openai-gpt-6-2-astra-sol-release/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/openai-gpt-6-2-astra-sol-release/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Sofia Almeida</dc:creator>
<description>OpenAI shipped GPT-6.2 Astra and Sol on November 17, with higher benchmark scores at lower prices — and Astra is the company’s first model rated High for AI self-improvement.</description>
<enclosure url="https://aifrontierpost.com/articles/openai-gpt-6-2-astra-sol-release/cover.webp" type="image/webp"/>
</item>
<item>
<title>Ship a support agent on your own site: ChatKit and the OpenAI Agents SDK</title>
<link>https://aifrontierpost.com/articles/chatkit-agents-sdk-support-agent-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/chatkit-agents-sdk-support-agent-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Priya Nair</dc:creator>
<description>Agent Builder shuts down November 30, 2026. The replacement is a stack you own: ChatKit for the chat UI, the Agents SDK for the brain, and a Python server you run yourself — tools, guardrails, and all.</description>
<enclosure url="https://aifrontierpost.com/articles/chatkit-agents-sdk-support-agent-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>Design in your coding agent: hands-on with Google&#x27;s Stitch skills</title>
<link>https://aifrontierpost.com/articles/stitch-skills-google-stitch-design-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/stitch-skills-google-stitch-design-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Marcus Doyle</dc:creator>
<description>Google&#x27;s Stitch skills put a UI design studio inside Claude Code, Cursor, and Codex: 16 skills, three plugins, and one MCP server between a text prompt and reviewed React components. This tutorial installs them, connects the MCP server, and runs the full loop — every command verified.</description>
<enclosure url="https://aifrontierpost.com/articles/stitch-skills-google-stitch-design-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>Corma hits a $1 billion valuation three months out of stealth — the bet that AI defense needs its own models</title>
<link>https://aifrontierpost.com/articles/corma-125m-series-a-1b-valuation/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/corma-125m-series-a-1b-valuation/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Priya Nair</dc:creator>
<description>Corma raised $125 million at a $1 billion valuation three months after leaving stealth, betting that cybersecurity defense deserves its own foundation models — not repurposed general-purpose ones.</description>
<enclosure url="https://aifrontierpost.com/articles/corma-125m-series-a-1b-valuation/cover.webp" type="image/webp"/>
</item>
<item>
<title>Turn lab notes into a submission-ready paper: hands-on with Google Research&#x27;s PaperOrchestra</title>
<link>https://aifrontierpost.com/articles/paper-orchestra-google-research-paper-writing-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/paper-orchestra-google-research-paper-writing-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Priya Nair</dc:creator>
<description>Google Research&#x27;s PaperOrchestra turns an idea and an experimental log into a submission-ready LaTeX manuscript with five specialised agents. It gained 1,175 stars in a day — this tutorial installs it, feeds it real materials, and shows exactly where the free tier ends.</description>
<enclosure url="https://aifrontierpost.com/articles/paper-orchestra-google-research-paper-writing-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>Gemini 4 Hydrogen gets hands: Google&#x27;s Antigravity robotics plugin turns plain conversation into robot control</title>
<link>https://aifrontierpost.com/articles/gemini-4-hydrogen-antigravity-robotics-plugin/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/gemini-4-hydrogen-antigravity-robotics-plugin/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Marcus Doyle</dc:creator>
<description>Google DeepMind posted a 31-second clip of a robot arm working from plain-English instructions — and launched a Robotics plugin that turns Antigravity into a workbench for going from simulation to hardware in about an hour.</description>
<enclosure url="https://aifrontierpost.com/articles/gemini-4-hydrogen-antigravity-robotics-plugin/cover.webp" type="image/webp"/>
</item>
<item>
<title>Microsoft launches the AI Transparency Initiative: 40+ companies to set open standards for governing AI agents</title>
<link>https://aifrontierpost.com/articles/microsoft-ai-transparency-initiative-ignite-2026/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/microsoft-ai-transparency-initiative-ignite-2026/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Marcus Doyle</dc:creator>
<description>Microsoft unveiled the AI Transparency Initiative at Ignite, a joint effort with more than 40 companies to build open standards so enterprises can see what their AI agents do, verify it independently, and stay in control when models change.</description>
<enclosure url="https://aifrontierpost.com/articles/microsoft-ai-transparency-initiative-ignite-2026/cover.webp" type="image/webp"/>
</item>
<item>
<title>Run MBZUAI&#x27;s K2 Horizon on your own machine: the fully-open fleet from 0.9B to 375B</title>
<link>https://aifrontierpost.com/articles/k2-horizon-local-run-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/k2-horizon-local-run-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Sofia Almeida</dc:creator>
<description>MBZUAI&#x27;s K2 Horizon is a fully open fleet of six reasoning models, from 0.9B to 375B parameters — weights, training data, code, and checkpoints included. Run the 0.9B on a laptop CPU in ten minutes, then scale to the 375B flagship with the same API.</description>
<enclosure url="https://aifrontierpost.com/articles/k2-horizon-local-run-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>Perplexity crosses a $1 billion revenue run rate — and pulls its IPO forward to early 2027</title>
<link>https://aifrontierpost.com/articles/perplexity-1-billion-revenue-run-rate/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/perplexity-1-billion-revenue-run-rate/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Priya Nair</dc:creator>
<description>Perplexity has crossed a $1 billion annual revenue run rate, CEO Aravind Srinivas said onstage at Semafor’s Silicon Valley &amp; The World event — up from under $250 million in January. The company says it got there burning a fifth of Anthropic’s cash, and it may now file for an IPO as early as the first quarter of 2027.</description>
<enclosure url="https://aifrontierpost.com/articles/perplexity-1-billion-revenue-run-rate/cover.webp" type="image/webp"/>
</item>
<item>
<title>Turn Claude Code into a bug bounty hunter: hands-on with Claude-BugHunter</title>
<link>https://aifrontierpost.com/articles/claude-bughunter-bug-bounty-hands-on-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/claude-bughunter-bug-bounty-hands-on-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Sofia Almeida</dc:creator>
<description>Claude-BugHunter is the number-one trending repo on GitHub: 83 Claude Code skills that turn a coding agent into a methodical bug bounty hunter. Install it, run a practice hunt on Juice Shop, and put your first finding through the 7-question validation gate — every deterministic step verified.</description>
<enclosure url="https://aifrontierpost.com/articles/claude-bughunter-bug-bounty-hands-on-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>Higgsfield hands-on: Soul ID keeps one face consistent from photo to cinematic video</title>
<link>https://aifrontierpost.com/articles/higgsfield-soul-id-cinema-studio-hands-on-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/higgsfield-soul-id-cinema-studio-hands-on-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Sofia Almeida</dc:creator>
<description>Train a Soul ID from 20–80 photos for 25 credits, generate the portrait in Soul 2.0, then animate it in Cinema Studio with a real camera move — the same face, every scene.</description>
<enclosure url="https://aifrontierpost.com/articles/higgsfield-soul-id-cinema-studio-hands-on-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>Talk to Blender: build 3D scenes with Claude and MCP for Blender</title>
<link>https://aifrontierpost.com/articles/mcp-for-blender-ai-3d-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/mcp-for-blender-ai-3d-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Sofia Almeida</dc:creator>
<description>MCP for Blender wires your LLM into Blender over a local socket: install the server, point your coding agent at it, and describe the scene you want. Every command below was run and verified — including a real render.</description>
<enclosure url="https://aifrontierpost.com/articles/mcp-for-blender-ai-3d-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>Isomorphic Labs raises $5 billion to push AI-designed drugs toward human trials</title>
<link>https://aifrontierpost.com/articles/isomorphic-labs-5-billion-series-c-ai-drug-design/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/isomorphic-labs-5-billion-series-c-ai-drug-design/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Sofia Almeida</dc:creator>
<description>Isomorphic Labs has closed a $5 billion Series C co-led by Greenoaks and Sequoia Capital, just six months after its $2.1 billion Series B. The money funds its IsoDDE drug-design engine and the push to take its first AI-designed medicines into human trials — the test the whole field has been waiting for.</description>
<enclosure url="https://aifrontierpost.com/articles/isomorphic-labs-5-billion-series-c-ai-drug-design/cover.webp" type="image/webp"/>
</item>
<item>
<title>Train your own LLM from scratch: hands-on with MiniMind&#x27;s 64M-parameter model</title>
<link>https://aifrontierpost.com/articles/minimind-train-tiny-llm-from-scratch-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/minimind-train-tiny-llm-from-scratch-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Marcus Doyle</dc:creator>
<description>MiniMind trains a 64-million-parameter language model from scratch for about ¥3 of GPU rental. This tutorial runs the whole pipeline — inference, a tiny pretraining pass, and SFT — on an ordinary CPU, with every command verified.</description>
<enclosure url="https://aifrontierpost.com/articles/minimind-train-tiny-llm-from-scratch-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>Nvidia’s Vera Rubin sweeps every MLPerf Training v6.1 benchmark — but the headline 10x agentic gain is still provisional</title>
<link>https://aifrontierpost.com/articles/nvidia-vera-rubin-sweeps-mlperf-training-v6-1/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/nvidia-vera-rubin-sweeps-mlperf-training-v6-1/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Sofia Almeida</dc:creator>
<description>Nvidia’s Vera Rubin platform made its MLPerf debut on November 16, topping every Training v6.1 benchmark and posting up to twice the per-GPU performance of the GB300 systems it succeeds — while its eye-catching 10x agentic throughput-per-watt figure still awaits MLCommons verification.</description>
<enclosure url="https://aifrontierpost.com/articles/nvidia-vera-rubin-sweeps-mlperf-training-v6-1/cover.webp" type="image/webp"/>
</item>
<item>
<title>Run the new open-weight speech champion yourself: hands-on with Salesforce&#x27;s SOV-Transcribe-1</title>
<link>https://aifrontierpost.com/articles/sov-transcribe-1-hands-on-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/sov-transcribe-1-hands-on-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Priya Nair</dc:creator>
<description>Salesforce&#x27;s SOV-Transcribe-1 leads every downloadable transcriber on the Artificial Analysis benchmark at 2.28% word error — and it takes plain-language instructions for how the transcript should read. This tutorial runs it two ways: a one-shot Transformers script and a vLLM server with an OpenAI-compatible endpoint.</description>
<enclosure url="https://aifrontierpost.com/articles/sov-transcribe-1-hands-on-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>Turn your coding agent into a film crew: hands-on with OpenFilm, the open-source AI video agent</title>
<link>https://aifrontierpost.com/articles/openfilm-ai-video-agent-hands-on-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/openfilm-ai-video-agent-hands-on-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Marcus Doyle</dc:creator>
<description>OpenFilm turns your coding agent into a film crew: a film is an HTML file, each shot a web page with one function, frame(t), and a real editor lets you change anything. We installed 0.3.2, wrote a four-second film by hand, and rendered a real 1920×1080 MP4 — no account, no API key, no video model.</description>
<enclosure url="https://aifrontierpost.com/articles/openfilm-ai-video-agent-hands-on-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>DeepSeek starts rationing its free chatbot: daily quotas, 40,000-character prompts, and a compute shortage it can&#x27;t hide</title>
<link>https://aifrontierpost.com/articles/deepseek-free-chat-quotas-compute-shortage/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/deepseek-free-chat-quotas-compute-shortage/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Marcus Doyle</dc:creator>
<description>On November 16, DeepSeek&#x27;s free app and web chat began capping single prompts at about 40,000 characters, dropping week-old conversations, and cutting users off with a daily quota — citing “insufficient computing power” — while its API stays pay-as-you-go.</description>
<enclosure url="https://aifrontierpost.com/articles/deepseek-free-chat-quotas-compute-shortage/cover.webp" type="image/webp"/>
</item>
<item>
<title>Google: AI agents now turn security patches into working exploits with no human in the loop — and a North Korea-linked crew rigged Claude Code</title>
<link>https://aifrontierpost.com/articles/google-gtig-exploit-agent-sombermeme/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/google-gtig-exploit-agent-sombermeme/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Priya Nair</dc:creator>
<description>Google’s threat intelligence team says attackers have crossed a line it had not seen before: AI agent frameworks that read a security patch, write the exploit and fire it, with no human in the loop. The same report describes a suspected North Korean crew turning Claude Code itself into a backdoor delivery pipe.</description>
<enclosure url="https://aifrontierpost.com/articles/google-gtig-exploit-agent-sombermeme/cover.webp" type="image/webp"/>
</item>
<item>
<title>Build a fully local RAG app with LlamaIndex and llama.cpp — no API keys, no cloud</title>
<link>https://aifrontierpost.com/articles/llamaindex-local-rag-llamacpp-tutorial/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/llamaindex-local-rag-llamacpp-tutorial/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>Tutorials</category>
<dc:creator>Sofia Almeida</dc:creator>
<description>LlamaIndex 0.14.26 turns your own documents into a question-answering app in about 30 lines — with a local Qwen3 model and nomic embeddings, no API key, no GPU, $0. We ran it end to end on a CPU-only machine: 3/3 answers right, 4/5 top-1 retrievals, and a chat engine that remembers what &#x27;they&#x27; means.</description>
<enclosure url="https://aifrontierpost.com/articles/llamaindex-local-rag-llamacpp-tutorial/cover.webp" type="image/webp"/>
</item>
<item>
<title>China&#x27;s top AI labs are shipping new models with the hacking skills cut out</title>
<link>https://aifrontierpost.com/articles/chinese-ai-labs-trim-hacking-skills-new-models/</link>
<guid isPermaLink="true">https://aifrontierpost.com/articles/chinese-ai-labs-trim-hacking-skills-new-models/</guid>
<pubDate>Tue, 17 Nov 2026 00:00:00 +0000</pubDate>
<category>AI News</category>
<dc:creator>Marcus Doyle</dc:creator>
<description>Moonshot, Tencent and Alibaba have all shipped recent models with offensive cyber capabilities deliberately trimmed, Bloomberg reports — and Zhipu plans to gate its most sensitive functions behind verification. The shift follows Beijing&#x27;s proposed export ban on offensive cyber AI.</description>
<enclosure url="https://aifrontierpost.com/articles/chinese-ai-labs-trim-hacking-skills-new-models/cover.webp" type="image/webp"/>
</item>
</channel>
</rss>