<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel>
<title>Agentic AI · Özer Cem KELAHMET</title>
<link>https://www.ozercemkelahmet.com/topics/agentic-ai</link>
<atom:link href="https://www.ozercemkelahmet.com/feeds/agentic-ai.xml" rel="self" type="application/rss+xml"/>
<description>New articles on Agentic AI from www.ozercemkelahmet.com</description><language>en</language>
<item><title>Performance Factory: how OpenAI checks the agents that debug ChatGPT in production</title><link>https://www.ozercemkelahmet.com/articles/agents-investigating-production-performance-factory</link><guid>https://www.ozercemkelahmet.com/articles/agents-investigating-production-performance-factory</guid><pubDate>Sat, 21 Nov 2026 07:00:00 +0000</pubDate><description>OpenAI runs Codex agents on ChatGPT slowdowns. How durable task state, adversarial review, leak-free backtests and a synthetic Google web keep them honest.</description></item>
<item><title>Patch-to-exploit at machine speed: Google&#x27;s exploit_agent and shorter disclosure windows</title><link>https://www.ozercemkelahmet.com/articles/patch-to-exploit-at-machine-speed</link><guid>https://www.ozercemkelahmet.com/articles/patch-to-exploit-at-machine-speed</guid><pubDate>Wed, 18 Nov 2026 07:00:00 +0000</pubDate><description>Google found an attacker tool that turns security patches into exploits with no human involved. Anthropic is cutting disclosure windows. Here is what it means.</description></item>
<item><title>Measuring AI that builds AI: Singularity-Bench and Anthropic&#x27;s three-day AMD kernel port</title><link>https://www.ozercemkelahmet.com/articles/measuring-ai-that-builds-ai</link><guid>https://www.ozercemkelahmet.com/articles/measuring-ai-that-builds-ai</guid><pubDate>Wed, 18 Nov 2026 07:00:00 +0000</pubDate><description>Singularity-Bench scores agents against human baselines on six AI-building tasks, while Anthropic claims Mythos ported Claude to AMD GPUs in three days.</description></item>
<item><title>Agentic AI frameworks explained: the Lang family and every major alternative</title><link>https://www.ozercemkelahmet.com/articles/agentic-ai-frameworks-langchain-langgraph-and-every-alternative</link><guid>https://www.ozercemkelahmet.com/articles/agentic-ai-frameworks-langchain-langgraph-and-every-alternative</guid><pubDate>Wed, 18 Nov 2026 07:00:00 +0000</pubDate><description>Every agent framework that matters in November 2026: LangChain, LangGraph, LangSmith, Langflow, Langfuse and the rest, compared with code and a guide to choosing.</description></item>
<item><title>AI agents get a front door and a wallet: Poppy 1.0, ChatGPT Wallet and Uber&#x27;s intent checks</title><link>https://www.ozercemkelahmet.com/articles/personal-agent-protocol-wallet-intent-checks</link><guid>https://www.ozercemkelahmet.com/articles/personal-agent-protocol-wallet-intent-checks</guid><pubDate>Tue, 17 Nov 2026 07:00:00 +0000</pubDate><description>How Sierra&#x27;s Poppy 1.0, OpenAI&#x27;s ChatGPT Wallet and Uber&#x27;s intent checks fit together: discovery files, OAuth delegation, UCP checkout and tokenized payments.</description></item>
<item><title>Gemini 4 Carbon tops MathArena: how to read scores, costs and intervals on a math leaderboard</title><link>https://www.ozercemkelahmet.com/articles/gemini-4-carbon-matharena-leaderboard</link><guid>https://www.ozercemkelahmet.com/articles/gemini-4-carbon-matharena-leaderboard</guid><pubDate>Tue, 17 Nov 2026 07:00:00 +0000</pubDate><description>Gemini 4 Carbon leads MathArena at 85.4%, but the top three overlap and costs differ 15-fold. How IRT, four runs, fresh problems and Lean shape the ranking.</description></item>
<item><title>Building an agentic AI system from scratch with LangGraph, Langfuse, DeepEval and RAGAS</title><link>https://www.ozercemkelahmet.com/articles/build-agentic-ai-from-scratch-langgraph-langfuse-deepeval-ragas</link><guid>https://www.ozercemkelahmet.com/articles/build-agentic-ai-from-scratch-langgraph-langfuse-deepeval-ragas</guid><pubDate>Tue, 17 Nov 2026 07:00:00 +0000</pubDate><description>Step by step: build a tool-using RAG agent with LangGraph, trace it with Langfuse, test it with DeepEval and score retrieval with RAGAS, with a live simulation.</description></item>
<item><title>Subagents become the default in coding agents: Antigravity, OpenCode and Copilot CLI</title><link>https://www.ozercemkelahmet.com/articles/subagent-orchestration-coding-agents</link><guid>https://www.ozercemkelahmet.com/articles/subagent-orchestration-coding-agents</guid><pubDate>Mon, 16 Nov 2026 07:00:00 +0000</pubDate><description>Antigravity /teamwork, OpenCode 2.3.0 and Copilot CLI 1.0.111 make subagents the default. How orchestrator-worker agents, context, budgets and permissions work.</description></item>
<item><title>METR&#x27;s AI evaluations explained: the 19% developer slowdown study, time horizons and Anthropic</title><link>https://www.ozercemkelahmet.com/articles/metr-ai-evaluations-developer-slowdown-study</link><guid>https://www.ozercemkelahmet.com/articles/metr-ai-evaluations-developer-slowdown-study</guid><pubDate>Mon, 16 Nov 2026 07:00:00 +0000</pubDate><description>What METR is, how it tests frontier AI, why its 2025 trial found developers 19% slower with AI, what time horizons measure and how its Anthropic report fits.</description></item>
<item><title>Generative AI in R&amp;D in 2026: where it measurably speeds up research and where it does not</title><link>https://www.ozercemkelahmet.com/articles/generative-ai-in-r-and-d-2026</link><guid>https://www.ozercemkelahmet.com/articles/generative-ai-in-r-and-d-2026</guid><pubDate>Mon, 16 Nov 2026 07:00:00 +0000</pubDate><description>Where generative AI measurably speeds up R&amp;D in 2026, from AI labs and drug discovery to chip design, how to measure the gains honestly and where to start.</description></item>
<item><title>Companies cap AI spend per employee while model prices fall: a practical guide to AI FinOps</title><link>https://www.ozercemkelahmet.com/articles/per-employee-ai-budgets-and-finops</link><guid>https://www.ozercemkelahmet.com/articles/per-employee-ai-budgets-and-finops</guid><pubDate>Tue, 10 Nov 2026 07:00:00 +0000</pubDate><description>Apple and Accenture reportedly capped staff AI spending as Anthropic, Meta and Fireworks cut prices. How to measure agent costs and set budgets that work.</description></item>
<item><title>OpenAI&#x27;s Control Room and the case for a real kill switch on AI workloads</title><link>https://www.ozercemkelahmet.com/articles/ai-workload-kill-switches-control-room</link><guid>https://www.ozercemkelahmet.com/articles/ai-workload-kill-switches-control-room</guid><pubDate>Tue, 10 Nov 2026 07:00:00 +0000</pubDate><description>OpenAI built Control Room to stop research AI jobs within 35 seconds. How workload inventories, halts at each layer, safety cases and stop drills work.</description></item>
<item><title>METR finds AI roughly doubling Anthropic&#x27;s research pace as its IPO filing flags a threshold</title><link>https://www.ozercemkelahmet.com/articles/ai-speeding-up-ai-research</link><guid>https://www.ozercemkelahmet.com/articles/ai-speeding-up-ai-research</guid><pubDate>Tue, 10 Nov 2026 07:00:00 +0000</pubDate><description>METR estimates Anthropic&#x27;s capability progress has roughly doubled, mostly because of AI help. How AI R&amp;D acceleration is measured and why thresholds matter.</description></item>
<item><title>Governing AI agent access with scoped credentials and undo buttons</title><link>https://www.ozercemkelahmet.com/articles/governing-ai-agent-access</link><guid>https://www.ozercemkelahmet.com/articles/governing-ai-agent-access</guid><pubDate>Sun, 08 Nov 2026 07:00:00 +0000</pubDate><description>A reported Palo Alto deal, Goldman&#x27;s tool limits and a deleted syllabus show why AI agent credentials and permissions are becoming their own security category.</description></item>
<item><title>Always-on AI agents move from chat sessions to background teammates</title><link>https://www.ozercemkelahmet.com/articles/always-on-ai-agents</link><guid>https://www.ozercemkelahmet.com/articles/always-on-ai-agents</guid><pubDate>Sun, 08 Nov 2026 07:00:00 +0000</pubDate><description>Perplexity, Meta, DeepSeek and Airbnb moved agents towards persistent background work this week. How long-running agents are built, governed and budgeted.</description></item>
<item><title>GPT-6 Cyber, Tencent Hy4 and DeepMind&#x27;s RSA work on AI bug hunting</title><link>https://www.ozercemkelahmet.com/articles/ai-models-that-find-vulnerabilities</link><guid>https://www.ozercemkelahmet.com/articles/ai-models-that-find-vulnerabilities</guid><pubDate>Sun, 08 Nov 2026 07:00:00 +0000</pubDate><description>OpenAI, Tencent and Google DeepMind all shipped security-focused AI work this week. What happened, how AI bug hunting works, and what defenders should do now.</description></item>
<item><title>Retrieval-augmented generation (RAG): from naive to agentic</title><link>https://www.ozercemkelahmet.com/articles/rag-from-naive-to-agentic</link><guid>https://www.ozercemkelahmet.com/articles/rag-from-naive-to-agentic</guid><pubDate>Tue, 20 Oct 2026 07:00:00 +0000</pubDate><description>How retrieval-augmented generation works, why naive RAG fails, and how advanced, graph, and agentic RAG fix it, with evaluation.</description></item>
</channel></rss>