Hermes Agent. 4 releases in 3 weeks.
Android support, GPT-5.5, 19 platforms including Teams, agent swarms, and a self-maintaining skill library that prunes itself every 7 days.
OpenClaw lost its founder to OpenAI. The gap is showing.
Full step-by-step install guide dropping in Sunday's newsletter.
McKinsey studied 6,800 skills to find what AI can replace.
Result: SQL, coding tests, invoicing → easiest to automate.
Leadership, negotiation, coaching → hardest to replace.
Companies hired for the wrong skills. For years.
The problem was never talent. It was how talent was measured.
Follow Jargonese for more.
SpaceX offered $60B for Cursor — and it's not really about the editor.
Cursor's edge isn't the AI model. It's the retrieval pipeline: Tree-sitter parses code into logical units, Merkle trees track only what changed, vectors search by meaning not keywords. Your raw code never leaves your machine.
The real story: the interface layer between devs and AI is the most defensible position in the stack.
Models commoditize. Workflows don't.
#AI #DeveloperTools #Cursor
Two Anthropic designers dropped their real Claude Design workflows.
I'm putting the full playbook which I found in my newsletter tomorrow
Best hack: handoff is a URL now, not a Figma file. Claude Code reads it directly and starts building.
Subscribe to get it → https://jargonese.beehiiv.com/subscribe
Murat Bayral and Can Hoskan ran a fair test: Claude Design vs Lovable, identical brief, one prompt, no follow-up.
They built a DESIGN.md file with full visual rationale, dropped the same brief into both tools, and compared outputs.
Claude Design (Opus 4.7) returned a full redesign in under 10 minutes — shaped by the context they gave it. Lovable was faster to start but leaned more on templates.
@sidj79
If you accept that all AI content you create would be traceable, your idea would work. There is no technical barrier preventing SynthID from, in the near future, also embedding your user ID into your videos. This looks like something the New World Order would want to pass as a mandate.
Santiago sold his company after 16 years and built an agent to handle his job search: Career-Ops, open source on GitHub.
740+ offers evaluated. 354 tailored CVs generated. 68 applications sent. Outcome: Head of Applied AI.
Crucially — it's HITL by design. The system automates analysis. The human makes every submission decision. Not spray-and-pray.
Built on Claude Code + Playwright. 12K stars in two days.
github.com/santifer/career-ops
Google’s Gemma 4 release is a strong signal for open AI
You now have:
A 31B model with GPT-4–level reasoning benchmarks
Multimodal capabilities
Long context (up to 256K tokens)
Local deployment options (even on consumer hardware)
Apache 2.0 license
This shifts control back to developers:
no API dependency, better privacy, and full customization
It’s not perfect — but it’s a meaningful step forward for open models
👉 I’ll share how to set it up with Claude Code next(Sunday's Newsletter)
You cropped the AI image, compressed it to 25%, and removed the watermark.
Google STILL knew it was AI-generated. 👀
It's called SynthID — a hidden signature embedded at the pixel level that survives editing, compression, and cropping.
Tested it myself. Edited an AI image 6 different ways. Uploaded to Gemini. Flagged instantly.
It works on videos and audio too.
With AI product photos flooding online shopping, this kind of transparency tool is genuinely needed.
Claude can make full product demo videos from a single prompt
Remotion handles the motion graphics. ElevenLabs adds the voiceover.
Zapier auto-publishes after approval.
No editing. No coding.
Just describe what you want.
The GitHub repo to set this all up yourself drops in tomorrow's newsletter — link in bio to subscribe
Earl