AI News Today | Julian Goldie Podcast

Hermes Agent OS Just Changed AI Agents Forever!

Build a 24/7 Agent Operating System with Hermes + Claude (Brain, Mission Control & Memory)The script demonstrates an Agent OS built with Hermes and Claude that turns a standalone AI agent into a “24/7 AI employee” by adding three layers: the brain (Hermes plus the models it uses, including optional free/local models), mission control (a single dashboard with buttons for chat, voice, agent teams, Kanban-style tasking, and a workspace where builds are saved), and a memory layer (an Obsidian notes vault agents read/write so sessions start with full context). It contrasts fragmented multi-tab workflows with a unified system where models are swappable without losing context, history, or running jobs. Examples include custom workflows for video creation, a voice agent that can control the computer, an “Oracle” that monitors news and competitors, drafts and publishes WordPress posts, and automated SEO/content production. The system and install zip are offered via the AI Profit Boardroom with tutorials, community support, and coaching.

Transcribe →

NEW DeepSeek V4 Flash Update is INSANE!

Yesterday00:10:30Tap to summarize

DeepSeek V4 Flash API Just Dropped (Built for AI Agents): Benchmarks, 1M Context & Hermes SetupDeepSeek-V4 Flash has officially moved from preview to a live API release aimed at AI agents, with major benchmark jumps across evaluations while keeping the same model architecture and size as the preview—improvements come from training, not scale. The script walks through what changed, how the model compares to other models like Opus 4.8, and why it’s fast and inexpensive while enabling up to a million tokens of context for longer coding loops and agentic work. The creator stress-tests it on Goldy Bench by building 50+ quick projects (including 2D/3D games) to evaluate planning, UI, and prompt handling, noting it’s not “frontier” level but delivers clean outputs. They demonstrate wiring V4 Flash into their Agent OS in minutes via a “DeepSeek Coder” tab and a Hermes Agent profile, explain switching models in Hermes, clarify the upgrade applies only to the V4 Flash API (not web/app or V4 Pro yet), and promote their AI Profit Boardroom/Agent OS resources, tutorials, and community.

Transcribe →

4 FREE Ways to Save 80% on Claude Claude Tokens!

Yesterday00:12:42Tap to summarize

Cut Claude Code Tokens by 80% (4 Free GitHub Repos + Full Token Minimization Stack)The video shares a free “token minimization playbook” that claims to cut Claude Code token usage by about 80% using four open-source GitHub repos—RTK, Caveman, Ponytail, and Omniroute—and notes they can work with any AI agent. RTK sits between the agent and shell commands to compact and filter tool output (tested at 82.9% reduction, adding ~14ms). Caveman removes fluffy, overly polite responses to reduce reply tokens (tested 69% fewer output tokens, 37% overall reduction). Ponytail reduces unnecessary code volume by avoiding speculative abstractions while keeping safety basics. Omniroute routes grunt work to free models via a local gateway with 93 models and built-in RTK/Caveman compression. The script also suggests /clear, /compact, trimming claude.md, routing by difficulty, batching requests, planning first, and using a scout sub-agent, and promotes getting the full setup inside the AI Profit Boardroom’s Agent OS.

Transcribe →

Sakana Fugu Ultra 1.1 DESTROYS Fable 5?

Yesterday00:12:48Tap to summarize

Sakana Fugu Ultra 1.1: Multi‑Agent “Council of Models” That Beats Fable 5 (Goldy Bench Tests)The video reviews Sakana Fugu Ultra 1.1, a new Japan-based model that benchmarks above Fable 5 and uses a “council of models” approach: one prompt is routed by an orchestrator to up to three expert models whose outputs are merged into a single build. The creator tests it on Goldy Bench by generating multiple one-shot game builds (e.g., open-world and flight simulator) to evaluate planning, logic, coding, and UI, noting the main drawback is slow generation (about 15–25 minutes) with no back-and-forth. Side-by-side comparisons show Fugu Ultra producing smoother, less buggy, better-looking results than Fable 5. The script also mentions regional blocking in the EU/UK, alternatives like OpenRouter Fusion and Hermes mixture-of-experts, and integrating everything into an Agent OS with saved workspace, plus access options via console.sakana.ai or OpenRouter and a pitch for the AI Profit Boardroom.

Transcribe →

Agent Operating Systems: Automate ANYTHING!

Yesterday00:09:39Tap to summarize

Agent OS Q&A: Hermes Wake Word, Free Claude Code Routers, Buzz, Goal Mode, and Building Custom Company AgentsThe episode answers recent AI Profit Boardroom community questions about the Agent Operating System, highlighting new features like a Hermes wake word, free Claude Code via OmniRoute and Nine Router, a free AI coder workspace with previews/history, and integrated memory tools like Gemini Notebooks and Gemini Flash. It explains how to build sub-agents for a large company (e.g., an HR agent) by using a shareable custom GPT trained on policies, with more advanced options like Buzz. The video discusses Pi as a customizable agent platform, demonstrates /Goal mode in Hermes and Codex, clarifies that Fable 5 remains in Claude Code, and compares Hermes vs Claude usage and customization. It defines Buzz as Slack-like agent coordination, shares a member’s automated SEO machine build, suggests edtech uses like voice tutoring and automated course video creation, and recommends an iterative build process for customizing Agent OS modules, plus resources on running it free and minimizing tokens.

Transcribe →