AI Valley
Posts
The OpenAI Files: A rare document

The OpenAI Files: A rare document

PLUS: Midjourney launches its first AI video generation model, V1

Barsee
19 Jun

_{Sign up}_|_{Follow us on X}_|_Sponsor

Together with

Howdy! It’s Barsee again.

Happy Thursday, AI family, and welcome back to AI Valley.

Today’s climb through the Valley reveals:

Midjourney launches its first AI video generation model, V1
Meta’s $100 million AI job offer to OpenAI’s employees
The OpenAI Files: A rare document
AI learns to reason without labels
Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

WISPR FLOW

Voice-First Writing Superpower 🗣️📱

Image: Flow

Speak ideas once, and Flow instantly turns them into polished, typo-free text in 100+ languages. Natural voice, AI rewrites, and auto-formatting that’s 3× faster than typing.

Here’s how to do it:

Download Wispr and sign up.
Enable Wispr Flow Keyboard:
Settings → Keyboard → Add Wispr Flow.
Hold 🌐 → Select Wispr Flow.
Use anywhere:
Open any app → Tap 🌐 → Tap 🎙️ → Start speaking.
You can turn on Casual Mode and speak in a natural way.
Settings → Flow → Casual tone.
Wispr will auto-format your speech:
1. Removes filler words
2. Capitalizes names
3. Rewrites mid-sentence edits into clear messages

Download Wispr Flow and stop wasting time typing what you could just say.

_{*This is sponsored}

MIDJOURNEY

Midjourney launches its first AI video generation model, V1 🎥

Video generated with V1 (by AIBreakfast)

Midjourney, one of the most popular AI image generation startups, launches its much-anticipated AI video generation model, V1.

Here's what you need to know:

Unlike OpenAI’s Sora or Kuaishou’s Kling 2, V1 doesn’t support text-to-video (yet). You’ll need to generate still images with Midjourney’s V7 model (or upload your own) before hitting “Animate” to produce four 5-second video variants.
True to its roots, V1 remains exclusive to Discord (no web or mobile app yet).
Videos cost 8x more than images. On the $10/month Basic plan, you’ll get roughly 10 video generations. Pro ($60) and Mega ($120) subscribers get unlimited generations in “Relax” mode, but it’s slower.
CEO David Holz warned that unpredictable GPU costs could force pricing changes in the coming weeks.
The launch coincides with ongoing lawsuits from Disney and Universal, alleging copyright misuse in Midjourney’s training data, highlighting examples like Darth Vader. V1’s early outputs often mirror iconic franchises and artists’ aesthetics.
Much like its image generation tools, V1’s videos lean more dreamlike than photorealistic. The first impressions are mostly positive from users, but it’s too early to call it a Sora competitor.

midjourney introduces video generation and it’s surpassing all my expectations.
— Phi Hoang (@apostraphi)
5:35 PM • Jun 18, 2025

Why it matters:

V1 marks Midjourney’s big leap into AI video, testing whether its devoted artist community is ready to pay premium prices for animated content. But legal risks loom large, and the Discord-only, credit-intensive setup may limit broader adoption. Still, with ambitions that hint at “real-time open-world simulations,” Midjourney seems to be playing a longer, more immersive game beyond just video content.

THROUGH THE VALLEY

1. Meta’s $100 million AI job offer to OpenAI’s employees

According to OpenAI CEO Sam Altman, Meta is aggressively trying to poach top AI talent, reportedly offering engineers up to $100 million in signing bonuses, though he says none of their best staff have accepted. Meta views OpenAI as a key rival and is racing to catch up, investing $14.3 billion for a 49% stake in Scale AI and recruiting its CEO, Alexandr Wang, to lead a new superintelligence lab. While Altman calls Meta’s tactics damaging to innovation, others cite its open-source Llama models and hires like DeepMind’s Jack Rae as signs it remains a major AI force.

2. MIT completed the first brain scan study of ChatGPT users

A visualization of a new study on AI chatbots by MIT Media Lab

MIT’s latest study paints a stark picture of how overusing ChatGPT could erode our brain’s critical thinking abilities. Over four months, researchers tracked 54 participants writing essays using ChatGPT, Google, or no tools at all, monitoring their brain activity with EEG scans. ChatGPT users had 47% less brain connectivity and couldn’t recall essays written just minutes earlier. Teachers described their work as “soulless” and lacking insight. When forced to write without AI, these users performed worse than those who never used it. The study warns of growing cognitive debt, a tradeoff between short-term efficiency and long-term brain function.

3. Software is changing (again)

Andrej Karpathy’s talk at AI Startup School outlines how software is evolving into a new era (Software 3.0) where we “program” using natural language instead of code. LLMs now serve three roles: utility (always-on services), fab (infrastructure-heavy like chip manufacturing), and OS (orchestrating compute and ecosystem wars). Unlike past tech that started with experts, LLMs are already in the hands of everyday users. Key takeaways: everyone’s a programmer now, LLMs need human oversight, autonomy will be adjustable, and future software should be built for agents. If you’re not thinking in prompts yet, you’re falling behind.

BEST AI LINKS

Trending Tools

Generated Assets: Turn any idea into an investable index using Generated Assets. Analyze performance, compare to the S&P 500, fine-tune holdings, and share it with the world. *
AgentX: Build a team of AI agents that work together with any LLM.
Wonderish: Canva of Vibe Coding.
Alta: Your personal AI stylist that truly gets you.
Audio Overview: Listen to your Google Search results.

Think Pieces and Brain Boost

Coding AI agents have crossed a chasm.
Why does Claude Code feel like magic?
What happens when you feed AI nothing?
A person created a prompt that stops ChatGPT from hallucinating.

Science and Futurology

Amazon’s Zoox opens its first major robotaxi production facility.
Sam Altman says OpenAI has new self-driving car technology that is better than any other current approach.

Miscellaneous

24 US AI startups that have raised $100M or more in 2025.
Six-month-old, solo-owned vibe coder Base44 sells to Wix for $80M cash.

_{* asterisk is sponsored content}

OPENAI

The OpenAI Files: A rare document 📂

Credit: Midas Project and Tech Oversight Project

After a year of research, watchdog groups Midas Project and Tech Oversight Project have released the OpenAI Files, a 50-page, 10,000-word interactive report packed with data visuals that track how OpenAI’s corporate structure has changed over time.

Here’s what you need to know:

The report centers on three major concerns: (1) Governance breakdown following the removal of OpenAI’s 100x profit cap, (2) Potential conflicts of interest tied to Sam Altman’s personal investments in rival startups, and (3) A pattern of rushed safety checks and what insiders call a “reckless” internal culture.
It charts OpenAI’s shift from a capped-profit nonprofit to a high-stakes commercial enterprise, spurred by a $13 billion deal with Microsoft. Internal records suggest investors directly pressured leadership to eliminate safeguards around AGI profit distribution.
The report details how Sam Altman was almost removed as CEO in 2023 after senior staff accused him of being dishonest. It also shows how Altman’s personal investment fund is closely tied to the direction OpenAI is taking with its products.
Commercial priorities reportedly led to mass content scraping without permission, minimal pre-launch safety vetting, and aggressive infrastructure scaling that triggered power grid failures in multiple local communities.

Why it matters:

These documents offer rare insight into how OpenAI’s founding mission (developing AGI for humanity’s benefit) has been steadily reshaped by investor influence and unchecked growth. With OpenAI refusing to comment, the findings highlight a growing case for external accountability as AGI development rapidly accelerates.