No slides, no hype. I teach by building live on YouTube and X: AI agents, local models, and AI workflows, mistakes included. Join live and ask questions in chat, catch the replays on YouTube, and follow on X for the latest AI updates.
people trained through talks, workshops, and mentoring
2.2k+
GitHub stars on the open-source tools I teach with
25+
years shipping software, the context behind every video
Workshops & talks
Bring an AI workshop to your school or organization.
The same way I teach on the livestreams, run for your students, faculty, or team: hands-on sessions on AI Agents, Local Models, and AI Workflows, in plain language. Online via Zoom, with on-site options on request.
Google said (Oct 28) Gemini 4 Carbon is its strongest frontier model, with cybersecurity and long-horizon work as the focus. It is not generally available. Starting this week, a subset of cyber defenders gets access, with a gradual rollout to more trusted partners through the Fairwind Program, while Google stays in the U.S. government’s voluntary pre-release testing process. Developers, businesses, and consumers come later, after more guardrail feedback. Google says Carbon ranks first on Artificial Analysis Intelligence Index v5.0 at 59, and reports 88.2% on SEC-bench Pro, 54.8% on ExploitGym, and 71.0% on Collinear’s CWE-bench. Those scores were measured without guardrails. On an internal benchmark, Google says secure-code success is more than 2× Gemini 4 Argon. It also says the Chrome team saw more than 10× better high-quality patches than another lab’s large frontier model, and that Wiz sees Carbon leading its web penetration-testing benchmarks. On WebDev Arena, Google says Carbon gained 89 Elo to 1768 and ranks fourth. Internally, Google says teams have let Carbon run for days on multi-step engineering tasks. Introductory API pricing matches Argon at $2 per million input tokens and $10 per million output tokens, with cached input 95% off. After the introductory period, Google says the price is $4 and $20. The post does not give the introductory end date. On safety, Google says successful cheating on Terminal-Bench 4.0 fell to 0.0% and cheat attempts to 0.4%, unsupported claims were cut about 2× versus Argon, overreach fell 26% and under-service 29%, and Gray Swan’s indirect prompt-injection attack success rate fell from 0.7% for Argon to 0.0% for Carbon. Those figures are Google’s. Distinct from the Gemini 4 Argon launch two weeks earlier.
GitHub said (Oct 28) Copilot Fusion is generally available. Developers pick Fusion in the model picker, then Standard, High, or Max effort, and Copilot routes and orchestrates models instead of the user choosing one model per step. GitHub says it combines Project HydraFusion’s multi-model orchestration with Copilot’s automatic model selection. Workflows it describes are a single model, a cascade that drafts with an efficient model and escalates after a quality check, and a critique loop in which one model drafts, another reviews, and the first revises. On the Artificial Analysis Coding Agent Index, GitHub says Fusion Max scored 60, within two points of Claude Code with Opus 5.5 Max, at about half the average API cost per task, about half as many tokens, and about 40% less time per task. Billing is each model’s standard rate. For a limited time, paid Copilot plans get a 30% discount on model costs while using Fusion. The JetBrains changelog says that discount lasts the first 30 days. Fusion is in Visual Studio Code, the Copilot app, Copilot CLI, and JetBrains IDEs. It only uses models an enterprise or organization has enabled, and admins can make Fusion the default. Research-preview selections of auto, HydraFusion, or HydraFusion Max move to the matching Fusion effort level. A same-day experimental preview can route Standard effort to supported on-device models, including a managed Windows download of MAI Code 1.1 Flash, or a manually configured MAI Code 1.1 Flash or Qwen3.8 Flash Next server. Local inference is not an offline session: routing stays in the cloud. Those scores and discounts are GitHub’s. Distinct from the earlier HydraFusion preview.
NVIDIA published (Oct 28) Nemotron 3.5 Transcribe, an open-weight speech model that follows a natural-language instruction for each request instead of locking one transcript style at training time. NVIDIA says it pairs a Transformer speech encoder and a frozen Nemotron 3 diarization model with the Nemotron 3.5 Lightning 30B-A3B hybrid Mamba–Transformer mixture-of-experts model, for about 4.3 billion active parameters. The released checkpoints are named NVIDIA-Nemotron-3.5-Transcribe-30B-A4B, in BF16, FP8, NVFP4, GGUF, and a DFlash2 draft model, under the OpenMDW-1.1 license, with a fine-tuning toolkit in NeMo Speech. NVIDIA says one model does verbatim, written-form, contextual, speaker-attributed, and rich transcription, speech translation, and spoken-language identification in 40 languages, on recordings up to one hour, and can return speaker-attributed word timestamps. It reports a mean word error rate of 4.71% on the nine public English Open ASR Leaderboard sets and 4.62% on the multilingual Open ASR Leaderboard, and says that is the best open-weight result on those boards. Given a list of expected names, NVIDIA says named-entity errors on ContextASR-Bench fall six-fold. With the Nemotron Transcribe NIM, it says the NVFP4 checkpoint plus DFlash2 transcribes up to 2,238 seconds of audio per second on one RTX PRO 6000 Blackwell GPU. Those scores and speeds are NVIDIA’s. Distinct from Nemotron 3.5 Bolt.
Source: NVIDIA Research
Updated October 28, 2026
What I teach: AI Agents, Local Models & AI Workflows
AI Agents
Autonomous agents with memory, tools, and skills that do the work, not just chat about it.
I teach AI Agents, Local Models, and AI Workflows to people who want to use AI, not just read about it. Before the channel: senior engineering roles at Standard Chartered Bank and Ohmyhome (Nasdaq), then a run of AI products built and shipped end to end. That's why every video, talk, and post is grounded in what actually ships.
Livestreams AI builds on YouTube and X
Speaker at NTU Singapore, PSIA, and DICT Philippines
300+ trained through workshops and mentoring
Open-source maintainer: Codex orchestrator, Local Evals, Agent Monitor
Subscribe on YouTube for the next livestream and follow on X for the latest AI updates. Want a workshop for your school or organization, or help with a product? Book a call.
