In partnership with
|
BUILDER HUSTLE
AI · TECH · STARTUPS
|
|
|
In This Issue
|
|
🧠 Gemini 3.5 Pro Rumored to Land With a 2 Million Token Context
|
|
🔀 Stanford's Shepherd Brings Git-Style Rollbacks to AI Agents
|
|
☂️ An Engineer Built a Drone That Follows You With an Umbrella
|
|
🧾 TaxHacker Auto-Extracts Your Receipts for Tax Season
|
|
|
🧠 GOOGLE · GEMINI 3.5 PRO
|
|
Rumor Mill Says Gemini 3.5 Pro Could Ship With a 2 Million Token Context.
|
|
|
Word going around is that Google's upcoming Gemini 3.5 Pro will launch with a 2 million token context window, roughly double what Anthropic's current top models reportedly support. If accurate, that would make it one of the largest context windows among frontier models, with real implications for working across huge codebases, long documents, and extended multi-turn conversations without losing earlier context.
|
|
📌 This is still an unconfirmed rumor, so treat the 2 million figure as directional rather than final until Google confirms it.
|
|
⚡ A larger context window mainly pays off in tasks like whole-repo code review or reasoning over very long documents in one pass.
|
|
|
7 Proven External Traffic Strategies
Most eCommerce brands running external traffic aren't scaling — they're just spending.
Wrong channels, no real attribution, and at the end of the month, still no clear answer to the question that matters: what actually drove revenue?
The brands getting it right aren't necessarily spending more. They've just stopped guessing. They know which channels pull weight on Amazon listings, which ones bleed budget, and why affiliate and creator traffic outperforms on ROI when it's set up correctly.
Levanta put together a free playbook breaking down 7 proven external traffic strategies — where each one works, where it falls apart, and what it takes to scale without it becoming a second job.
Inside you'll see how top brands are driving millions in off-Amazon revenue and why most channels underdeliver when brands don't know what to look for.
If you're serious about growing outside of PPC, this is worth 5 minutes.
|
🔀 STANFORD · SHEPHERD · AGENT INFRASTRUCTURE
|
|
Shepherd Gives AI Agents an Undo Button, Stanford Style.
|
|
|
Stanford researchers released Shepherd, described as an agent-native version of Git that turns an agent's execution into a reversible, traceable sequence of events instead of a flat message log. It lets you fork, branch, and roll back to any earlier point in an agent's run, using copy-on-write state restoration that is reportedly about five times faster than a Docker commit, with roughly 95 percent KV cache reuse when replaying steps. In practice, this means a long agent run that goes wrong mid-way does not have to be scrapped and restarted, you can just rewind to before the mistake happened.
|
|
📌 Treating an agent run as a Git-like trace is a genuinely different mental model from the flat chat logs most agent frameworks use today.
|
|
⚡ High KV cache reuse on replay is the detail that actually makes rollback cheap instead of just theoretically possible.
|
|
|
Check out Shepherd on GitHub →
|
|
|
☂️ JOHN TSE · DRONE UMBRELLA · VIRAL DEMO
|
|
This Umbrella Flies Above You and Follows You Around, Hands Free.
|
|
|
Canadian engineer John Tse, who runs the I Build Stuff channel, built a fully autonomous flying umbrella that hovers overhead and follows a person as they walk. It combines drone technology with a depth sensing camera and computer vision so it can detect and track its owner without any handheld controller, and Tse says the current prototype stays airborne for about 10 to 15 minutes per charge. It went viral, with plenty of people calling it a genuinely clever piece of personal tech, while others raised fair concerns about wind stability, spinning blades near someone's head, noise.
|
|
📌 Depth sensing plus computer vision is what lets it follow a person reliably instead of just hovering in a fixed spot.
|
|
🔭 Safety around spinning blades near the head is probably the single biggest hurdle before something like this could ever be sold commercially.
|
|
|
Watch the flying umbrella in action →
|
|
|
🧾 TAXHACKER · OPEN SOURCE · SELF-HOSTED
|
|
TaxHacker Turns a Pile of Receipts Into a Tax-Ready CSV.
|
|
TaxHacker is a self-hosted, open source AI accounting app that scans receipts, invoices, and PDFs, including handwritten text, and automatically pulls out the data you need for tax filing. It works across languages and currencies, including crypto, and lets you write custom prompts to extract specific fields the way you actually need them organized. Everything runs on your own infrastructure, and the app exports a clean, tax-ready CSV at the end, cutting out most of the manual sorting that usually eats up tax season.
|
|
📌 Handwritten text support pushes this well beyond basic OCR tools that only handle clean, printed receipts.
|
|
⚡ Self-hosting matters a lot here, since financial documents are exactly the kind of data most people do not want sitting on a third party server.
|
|
|
Get TaxHacker on GitHub →
|
|
Domain Names + Web and Email Hosting You Need
Still paying GoDaddy or Namecheap prices? Porkbun sells most domains at cost for low, transparent registration and renewal pricing with no nonsense. Get free features like WHOIS privacy and SSL certificates, plus real human support 24/7, 365 days a year. Save $1 on your next domain name now.
Stay sharp,
Builder Hustle
