The AI voice wars begin

A note from David: Learn Claude Code, Your Way

To the person reading this who wants to level up on Claude Code: I built this for you. 🎁

It’s an interactive guide for all skill levels on Claude Code best practices.

Paste my prompt (or upload the MD file) into Claude Chat, pick your level, and it walks you through each section step by step, tailored to how you actually learn.

Plus, a quiz after each section & at the end to hold you accountable.

The course covers:

1/ How Claude Code Actually Works
2/ Protecting Your Secrets (.env)
3/ Your First Session (/init & CLAUDE.md)
4/ Context Management
5/ Getting the Best Work Out of Claude (screenshots, plan & thinking mode)
6/ Managing Claude Over Time (escape, # shortcut, /compact, /clear)
7/ Extending Claude (custom commands, MCP servers, GitHub integration)
8/ Hooks and the Agent SDK

We’re entering an era where learning can, and should, be hyper-personalized to you.

Check out the full post on LinkedIn so you can get the MD file to start learning.

Good morning,

Voice AI is quietly becoming the next battleground, and open-source is leading the charge. At the same time, infrastructure breakthroughs and massive valuations signal where the real money is flowing.

Let’s dive in 👇

Want to engage in conversation about AI news? Join the Byte-Sized Reddit Community

🔊 Voice AI Is Breaking Open

🎙️ Cohere Drops Open Voice Transcription Model

Cohere launches an open source voice model focused specifically on transcription, signaling a push toward specialized, high-accuracy speech systems. Instead of general-purpose models, this approach prioritizes enterprise-grade reliability for meetings, calls, and documentation workflows. It also reinforces a broader shift where open-source models are catching up fast in production use cases.

🗣️ Mistral Enters Speech Generation

Mistral releases a new open source model aimed at generating speech, expanding beyond text into full multimodal interaction. This puts them in direct competition with players building voice agents, assistants, and real-time conversational systems. The bigger trend is clear, voice is becoming a core interface layer, not just an add-on.

🤖 AI Teachers Show Up at the White House

Robot joins Melania Trump White House event highlighting AI-powered education tools, including robotic teaching assistants. The demonstration focused on personalized learning and scalable instruction through AI systems. It’s an early signal that AI in education is moving from theory into visible public adoption.

💰 Infrastructure, Memory & Big Money

⚖️ Harvey Hits $11B Valuation

Harvey confirms $11B valuation Sequoia triples down as legal AI continues attracting massive investor confidence. The company is building deeply specialized workflows for law firms, not generic copilots. This reinforces that vertical AI, not horizontal tools, is where defensibility and revenue are strongest.

🧠 Google’s Memory Compression Breakthrough

Can Google’s AI memory compression algorithm help address the growing RAM bottleneck in AI systems. The approach compresses model memory usage while maintaining performance, which could dramatically lower infrastructure costs. If it works at scale, this impacts everything from local inference to enterprise deployments.

🛠️ Tools of the Day

→ Littlebird - AI-powered research assistant for tracking trends and insights
→ Claude Mobile Work Tools - Bring Claude workflows directly into mobile productivity use cases
→ Pio - AI tool for simplifying workflows and automating repetitive tasks

⚡ Quick Hits

→ Deccan AI raises $25M to scale AI training workforce
→ Google launches Lyria 3 Pro music generation model
→ Reddit bots face new human verification requirements
→ Conntour raises $7M for AI video search

🧾 TLDR

Voice AI is rapidly becoming the next major interface, with open-source models from Cohere and Mistral accelerating adoption. Meanwhile, infrastructure innovations like memory compression could reshape cost structures across AI deployments. Capital continues flowing into vertical AI companies like Harvey, reinforcing that specialization is winning. Across the board, the stack is getting cheaper, faster, and more domain-specific.

Cheers,
David