Forkcast Today #1: financial-services, UI-TARS-desktop, agentmemory, hello-agents, easy-vibe

Episode Summary#

The debut episode of Forkcast Today covers five trending AI infrastructure projects. Features include Anthropic’s financial services toolkit for building AI applications in finance, ByteDance’s multimodal desktop agent for GUI automation, a persistent memory system for AI coding agents, DataWhale’s comprehensive agent-building tutorial, and an AI-powered programming education tool that makes learning to code more accessible.

Repositories Covered#

financial-services#

Anthropic’s official toolkit and reference implementation for building AI applications in the financial services domain. Provides domain-specific patterns, compliance-aware workflows, and integration blueprints for banking, insurance, and fintech use cases.

UI-TARS-desktop#

ByteDance’s open-source multimodal AI agent that automates GUI interactions — it sees the screen, controls the mouse, and operates browsers and desktop applications. Features MCP tool integration and a vision-language model for understanding UI elements.

agentmemory#

A persistent memory system for AI coding agents that maintains context across sessions. Works with Claude Code, Codex, Cursor, and Copilot without client-specific dependencies.

hello-agents#

A comprehensive Chinese-language tutorial series from DataWhale that teaches how to build AI agents from scratch. Covers agent architecture, tool use, memory systems, and multi-agent orchestration with hands-on examples. The most popular AI agent learning resource on GitHub.

easy-vibe#

An AI-powered programming education platform that makes learning to code more intuitive and accessible. Uses vibe-based interactions to lower the barrier for beginners entering software development.

rowboat#

Your AI assistant starts every conversation from zero — it doesn’t remember your meetings, emails, or decisions, so you waste time re-explaining context Local-first AI coworker with persistent knowledge graph stored as plain Markdown in Obsidian-compatible vault, integrates Gmail, Google Calendar, Fireflies meeting notes, 1,608 commits, supports local models via Ollama/LM Studio + hosted models, MCP server + Composio tool support, optional Deepgram (voice input) and ElevenLabs (voice output), Apache 2.0

chrome-devtools-mcp#

AI coding agents can write code but can’t see what happens in the browser — they’re blind to layout bugs, console errors, network failures, and performance issues 42+ tools across 10+ categories (input, navigation, debugging, performance, network, memory, extensions, emulation), built on Puppeteer, 47 releases, 829 commits, v0.25.0 (May 2026), integrates with 19+ MCP clients (Claude Code, Cursor, Copilot, Gemini CLI, Windsurf, etc.), Chrome 144+ for auto-connect, Node.js 20.19+, Apache 2.0

MasterDnsVPN#

Standard VPN protocols are trivially blocked by deep packet inspection — DNS tunneling tools exist but are slow with 59-byte header overhead ~5-7B header overhead (88% lower than DNSTT’s 59B, 71% lower than SlipStream’s 24B), up to 9x faster downloads than DNSTT, 6 encryption methods up to AES-256-GCM, 8 resolver balancing modes, custom ARQ protocol with 600-client/800-server windows, survived 70+ day internet blackout in Iran, 36 releases, 1,001 commits, 17 platform targets, MIT

supersplat#

Editing 3D Gaussian splats requires desktop software with GPU setup — there’s no fast browser-based tool for inspecting, cleaning, and optimizing splats before publishing Free browser-based editor, no download or install needed, WebGL + WebGPU rendering, 144 releases (v2.25.1 May 2026), 619 commits, multi-language localization support, MIT license

dive-into-llms#

LLM theory is everywhere but hands-on programming tutorials covering fine-tuning, jailbreaking, watermarking, and safety alignment in one place are rare 11 chapters: fine-tuning, prompt learning, knowledge editing, mathematical reasoning (distill mini-R1), text watermarking, jailbreak attacks, LLM steganography, multimodal models, GUI agents, agent safety, RLHF safety alignment (PPO), developed from SJTU courses by Prof. Zhuosheng Zhang, Huawei Ascend collaboration edition available, v0.1.0, 232 watching

agent-skills#

A production-grade engineering skills library for AI coding agents by Addy Osmani. Standardizes engineering practices and quality gates to make agents more professional and reliable across various development workflows.

9router#

AI coding tools burn through paid API credits fast — you hit rate limits, get locked out, or run up a bill just from tool-call overhead 40+ API providers, 100+ models, RTK compression saves 20-40% tokens by compressing tool outputs, Caveman Mode saves up to 65% output tokens, 3-tier auto-fallback (subscription→cheap→free), OpenAI-compatible /v1 endpoint on port 20128, Node.js 20+ with Next.js 16 + React 19 + Tailwind CSS 4, 554 commits, 44 releases, MIT

oracle-ai-developer-hub#

AI agents need persistent memory across sessions but stitching together separate vector, graph, and relational databases creates fragmented context 5 apps (including multi-agent Chain-of-Thought RAG and finance AI agent with unified memory core), 11 notebooks covering RAG to agent reasoning, 5 agent-memory notebooks using oracleagentmemory package, 3 progressive workshops (IR→RAG→agents→memory-augmented agents), 613 commits, supports OCI GenAI, OpenAI GPT-5, LangChain/LangGraph, pre-configured Codespaces, UPL 1.0

Watch#

• Video: https://www.youtube.com/watch?v=mhfO_J6jxBM
• Cover: https://i.ytimg.com/vi/mhfO_J6jxBM/hqdefault.jpg

Notes#

Transcript and notes will be added from Forkcast output artifacts.