S²³⁴ubscribe²³⁴ | Ben’s Bites Pro²³⁴ | Ben’s Bites News⁴¹
Daily Digest #487

Want to get in front of 100k AI enthusiasts? Work with us here^54†

Hello folks, here’s what we have today;

PICKS

1. Artifacts on Claude AI are now available to all,⁶⁸⁰ and on mobile too. Plus, a new feature: you can highlight a bit of text & then 'improve' or 'explain', s/o to Mike from our community for sharing this. The Anthropic team also shared a behind-the-scenes video³³⁷ of how they came up with the idea for artifacts.

2. Google released three new models in a day.¹⁹⁸

   • An improved version of Gemini 1.5 Flash (Exp 0827)

   • An even smaller version of Flash: Gemini 1.5 Flash 8B.

   • An experimental version of Pro: Gemini 1.5 Pro (Exp 0827).

   More deets on the models and how to access them in 🍿Our Summary¹³ (also below)

3. But what’s the deal with OpenAI?¹⁹⁷ Well, it has shown its 🍓“Strawberry”³²⁴ project to federal officials. The OG result of this project is slow and expensive, so OpenAI is rushing to launch a distilled (baby) version (hopefully this fall). The scoop though? The Information reports that the larger version of 🍓 is in use for generating synthetic data—to train OpenAI’s next-gen model, code-named Orion.

4. Quickly build voice-driven AI products with AssemblyAI's Speech-to-Text models. Experience near-human accuracy, low-latency performance, and advanced Speech AI capabilities with a simple, robust API for rapid deployment and scalability. Sign up today¹⁰⁶ to receive $50 in free API credits! (sponsor)

TOP TOOLS

• AssemblyAI²⁵⁹* - Multilingual Speech-to-Text API with Near-Human Accuracy.

• Tolan^484† - A friendly little alien you can talk to about whatever.

• Erp AI²⁸⁴ - AI layer that sits over your ERP, helping you with smarter decisions without distractions.

• Clockwise Prism³⁹⁷ - An intelligent calendar with scheduling that works like magic.

• Ragie³⁶¹ - Fully managed RAG-as-a-service for developers.

• Brighterway³⁰⁹ - AI-powered review of medical records.

Some hacky nuggets I found:

• Math Studio¹⁹⁵ - An app to quickly learn new math topics. (v. low accuracy though).

• Using LLMs for emails³²² beyond just write-email-for-me.

View more →¹⁰⁰
*sponsored

NEWS

• AI productivity hacks⁵⁶⁹ for busy professionals. ✨ New blog from us.

• Google Meet’s automatic AI note-taking³⁴⁸ is here.

• Achieve more with GenAI²⁷⁵ - A Masterclass with Ethan Mollick, Allie K. Miller, Don Allen III, and Manuel Sainsily.

• A new research from Google uses image models to generate a completely playable game²¹⁶ of Doom.

• The impact of AI on the call centre industry³³¹ in the Philippines.

• Nvidia launched NIM Agent Blueprints¹¹⁰ for enterprises to make their own AI.

• Cerebras is running Llama 3.1 8B at 1800 tokens per second.²⁶⁷ Related read: Inference is free and instant²⁶⁷ (I think LLMs will get smarter too).

View more →⁵⁴

Unclassifieds - short, sponsored links

• Generate more leads - start using A.I. to acquire more paying customers… learn how for free⁷⁹

QUICK BITES

Google's rolling out some fresh AI models¹⁹⁸, and they're not messing around. We're talking upgrades, new variants, and a whole lotta potential for developers to play with.

What is going on here?

Google's released three experimental Gemini models for developers to tinker with.

What does this mean?

Google's flexing its AI muscles with some serious upgrades:

• Gemini 1.5 Pro (Exp 0827) - Beefed-up version that's especially good at coding and handling complex prompts.

• Gemini 1.5 Flash (Exp 0827) - Got a significant glow-up, promising better performance across the board. Sully is impressed⁴⁸ based on its performance on his internal evals.

• Gemini 1.5 Flash-8B: A new, even smaller model that's packing a punch despite its size.

The gains on Flash are especially insane. It’s now at #6 on LMsys Arena⁴⁴, near GPT-4o-mini but comes with 2M context window. All these models are available in the Gemini API and Google AI Studio⁵⁰. For small use cases, the models are essentially free on both. Give them a spin.

ps: The 1.5 Pro and Flash models come with a date stamp (0827), so you know you're getting the freshest version. The older experimental version (Pro 0801) will be soon removed.

Why should I care?

Well, you’re getting truly small and cheap models without losing too much on the smartness level—and it’s coming from Google. They’re shipping fast and when a giant like Google moves fast, the momentum it carries is scary (for OpenAI, who’s gone to shipping waitlists).

Share this story¹³

Ben’s Bites Insights

We have 2 databases that are updated daily which you can access by sharing Ben’s Bites using the link below;

• All 10k+ links we’ve covered, easily filterable (1 referral)

• 6k+ AI company funding rounds from Jan 2022, including investors, amounts, stage etc (3 referrals)