OpenAI hits pause on Voice Mode

PLUS: New AI simulates 500M years of evolution

Welcome, AI enthusiasts.

After a buzzy May demo, it looked like ChatGPT was about to give the AI world a new voice — but a new delay is hitting the mute button on the hotly anticipated feature.

Is this a broader sign of a safety-first approach, or is something else giving the AI giant cold feet? Let’s dive deeper…

In today’s AI rundown:

  • OpenAI delays ChatGPT voice upgrade

  • New AI simulates 500M years of evolution

  • Create your own AI chatbot

  • Claude gets new ‘Projects’ upgrade

  • 5 new AI tools & 4 new AI jobs

  • More AI & tech news

Read time: 4 minutes



Image source: OpenAI

The Rundown: OpenAI just announced that the company is postponing the launch of ChatGPT’s hotly-anticipated Voice Mode feature demoed in May, citing the need for further safety tests and refinements.

The details:

  • OpenAI previously stated the Voice Mode would roll out within ‘a few weeks’ of May’s Spring Update, with access now expected for paid users by the Fall.

  • An Alpha test group was initially planned for late June, but is now being pushed back a month to allow for more safety checks.

  • The initial demo stirred drama following its ‘Sky’ model’s similarity to actress Scarlett Johansson, though OpenAI denied any wrongdoing.

  • On a lighter note, the company also separately released its ChatGPT desktop app for Mac users, allowing for broader integration across files, screen-sharing, and conversations.

Why it matters: While it’s disappointing after already missing the previous launch timeline, a delay is better than putting out an under-delivered product — particularly given the hype. But overpromising and underdelivering is a risky business, especially given OpenAI’s ever-increasing competition (see Anthropic’s latest developments below).


The Rundown: Ready to push LLMs past chatbots and start driving efficiency in everyday workflows? OctoAI’s latest tutorial unlocks how LLM-structured output can tackle critical, time-consuming tasks like JIRA ticketing.

In this tutorial, you’ll learn:

  • To quickly deliver time savings with LLM-powered Jira automation

  • How structured output resolves the need for overly clever prompting/parsing

  • To leverage JSON mode to unify and orchestrate disparate data and tools

  • How OctoAI’s optimized endpoints drives down inference costs

Read the blog, then sign up for OctoAI and get started with $10 in free LLM tokens.


Image source: EvolutionaryScale

The Rundown: Biotech startup EvolutionaryScale just introduced ESM3, a new AI model capable of designing and simulating new proteins in ways that mimic millions of years of natural evolution.

The details:

  • ESM3 is a frontier language model that can reason over protein sequences, structures, and functions simultaneously.

  • In tests, ESM3 generated a new fluorescent protein 58% different from other known structures — a leap that typically takes nature 500M years of evolution.

  • The largest version of ESM3 has 98B parameters, and was trained on approximately 1.8T tokens of biological data.

  • EvolutionaryScale was created by former META researchers, and has already secured $142M in seed funding from tech giants like Amazon and Nvidia.

Why it matters: ESM3 is bringing us closer to a future where we can ‘program’ biology as easily as we write computer code — potentially revolutionizing fields from healthcare to environmental science and unlocking completely new areas of knowledge across fields.


The Rundown: D-ID is an AI tool that allows you to create a fully customizable conversational AI avatar in just a few simple steps, no coding required.


  1. Head over to the D-ID website and log in/sign up to get free credits.

  2. Click "Create" and select "AI agent" to access the creation tool.

  3. Choose your AI agent's appearance from pre-generated avatars or upload your own and configure its name, language, voice, and knowledge sources.

  4. Select “Create Agent” once it’s ready, and start a conversation!


The Rundown: Why are Fortune 500 companies everywhere switching to Galileo Luna for their GenAI evaluations? So they can detect hallucinations, prevent prompt attacks, enforce data privacy, and more — all in real time.

With Luna, you can:

  • Achieve 18% higher accuracy, 11x faster speed, and 97% lower costs than GPT-3.5

  • Eliminate the need for ground truth data, saving time and resources

  • Customize evaluation models to your specific needs with easy fine-tuning


Image source: Anthropic

The Rundown: Anthropic just launched a new feature for its Claude AI assistant called ‘Projects’, allowing users to better organize files, chats, and info within a single location for more personalized conversations.

The details:

  • Projects allows users to curate documents, code, and more in a shared 200K context window, giving Claude quick access to organizational needs and context.

  • Users can share individual Projects, with a new activity feed available to collaborate with teammates across conversation snapshots.

  • Custom instructions (similar to ChatGPT) are also now available for each Project, allowing for quickly tailoring the AI’s responses.

Why it matters: While Anthropic’s models have consistently been at the top of the industry, the platform has lacked the additional customization features of ChatGPT (to varying degrees of success). With updates like Projects and Artifacts, Claude is gaining all the tools needed to be an enterprise AI powerhouse.


  • 🚀 Relay - AI-powered automation with a human in the loop

  • 📈 Fintwit - AI copilot for stock market analysis and investing

  • 🎥 Zabracat - Turn text into impactful videos in minutes

  • ✍️ Grant Orb - Write winning grant proposals with AI

  • 🔍 Jobright- Discover job opportunities tailored to your goals

  • 🔐 xAI - Security Operations Lead

  • 📈 Cresta - Sales Enablement Manager

  • 🛡️ Shield AI - Product Manager, Hivemind Foundations

  • 📊 Dataiku - Product Operations Manager


Toys R Us debuted a short promotional film at the 2024 Cannes Lions Festival, which was created by the retailer’s creative agency using early access to Sora.

ElevenLabs launched its Reader iOS app, enabling users to listen to podcasts, news, and more using the company’s AI-generated voices.

Etched secured $120M in funding for the development of its transformer-specific ‘Sohu’ chip, claiming an ‘order of magnitude’ of improvement in speed and affordability compared to Nvidia’s Blackwell chips.

Beehiiv announced the acquisition of AI website builder Typedream, with plans to upgrade the platform’s web builder, templates, analytics, and more.

Claude 3.5 Sonnet moved into the second overall spot in the Lmsys Chatbot Arena behind GPT-4o while taking the top ranking in specific categories like coding and hard prompts.

OpenAI is reportedly set to restrict access to its API and platform in China and other unsupported regions, with the new measures set to take effect on July 9.



Get your product in front of over 600k+ AI enthusiasts

Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.


How would you rate today's newsletter?

Vote below to help us improve the newsletter for you.

Login or Subscribe to participate in polls.

If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.

Join the conversation

or to participate.