Greg Brockman Tells the Real Story of OpenAI's Near-Death Experience
Wednesday, April 22, 2026 · 10 min
The must-listen of the week comes from The Knowledge Project, where Greg Brockman — OpenAI co-founder and President — gives what sounds like the most candid account yet of the 72-hour crisis that nearly destroyed the company. We've had plenty of post-mortems on the Sam Altman firing saga, but Brockman was at the absolute center of it, and hearing him reconstruct those days firsthand is a different thing entirely. If you've been waiting for the definitive insider episode on that inflection point, this is probably it — and it reportedly doesn't stop at the drama, covering the broader AGI race and where OpenAI sees itself heading with GPT-5 and beyond. Elsewhere, the AI security story is getting more interesting and more urgent simultaneously. The WSJ's Tech News Briefing flagged that Anthropic's Mythos tool has surfaced thousands of previously unknown cybersecurity vulnerabilities, which is both a remarkable proof-of-concept for AI-assisted security research and a five-alarm warning for anyone running unpatched software stacks. Meanwhile, First Round's In Depth podcast went deep with the co-founders of Artemis, an AI-native security platform that just emerged from stealth with $70M raised — framing their whole thesis around an 'AI vs. AI' arms race where human-speed defenses simply can't keep up with AI-generated attacks. The timing with the Mythos news is striking. Rounding things out, TechCrunch's Equity took a look at the fusion energy funding surge — private investment jumped from $10B to $15B in a matter of months — which is one of those 'weird but real' venture stories where investors have genuinely made peace with decade-plus timelines. Not every podcast story needs to be about LLMs, and this one's a useful reminder that patient capital is finding its footing in deep tech. The Brockman episode is the clear centerpiece this week. Block the time.
Listen
Show notes
The must-listen of the week comes from The Knowledge Project, where Greg Brockman — OpenAI co-founder and President — gives what sounds like the most candid account yet of the 72-hour crisis that nearly destroyed the company. We've had plenty of post-mortems on the Sam Altman firing saga, but Brockman was at the absolute center of it, and hearing him reconstruct those days firsthand is a different thing entirely. If you've been waiting for the definitive insider episode on that inflection point, this is probably it — and it reportedly doesn't stop at the drama, covering the broader AGI race and where OpenAI sees itself heading with GPT-5 and beyond. Elsewhere, the AI security story is getting more interesting and more urgent simultaneously. The WSJ's Tech News Briefing flagged that Anthropic's Mythos tool has surfaced thousands of previously unknown cybersecurity vulnerabilities, which is both a remarkable proof-of-concept for AI-assisted security research and a five-alarm warning for anyone running unpatched software stacks. Meanwhile, First Round's In Depth podcast went deep with the co-founders of Artemis, an AI-native security platform that just emerged from stealth with $70M raised — framing their whole thesis around an 'AI vs. AI' arms race where human-speed defenses simply can't keep up with AI-generated attacks. The timing with the Mythos news is striking. Rounding things out, TechCrunch's Equity took a look at the fusion energy funding surge — private investment jumped from $10B to $15B in a matter of months — which is one of those 'weird but real' venture stories where investors have genuinely made peace with decade-plus timelines. Not every podcast story needs to be about LLMs, and this one's a useful reminder that patient capital is finding its footing in deep tech. The Brockman episode is the clear centerpiece this week. Block the time.
In this episode
- Greg Brockman: Inside the 72 Hours That Almost Killed OpenAI — Fs
Greg Brockman: Inside the 72 Hours That Almost Killed OpenAI - Skip to main content - Skip to header right navigation - Skip to site footer [The Knowledge Project Podcast](https://fs.blog/knowledge-project-podcast/) # Greg Brockman: Inside the 72 Hours That Almost Killed OpenAI The AI race, the future of AGI, and the inside story of OpenAI. Greg Brockman is the co-founder and President of…
- Fusion doesn't have a normal startup timeline, and investors are fine with that — Techcrunch
# Fusion doesn't have a normal startup timeline, and investors are fine with that | TechCrunch Published: 2026-04-22T21:07:33+00:00 ## Summary Fusion energy has seen a surge in private investment in companies, with private investment increasing from $10 billion to $15 billion in just months. This discussion is discussed on the episode of TechCrunch's Equity podcast, with guest host Tim De…
- Why Software Updates Are Suddenly More Urgent - Tech News Briefing - WSJ Podcasts — The Wall Street Journal
# Why Software Updates Are Suddenly More Urgent - Tech News Briefing - WSJ Podcasts Published: 2026-04-21T07:01:00+00:00 Type: Opinion ## Summary A new AI tool, Anthropic's Mythos, has found thousands of cybersecurity risks and cybersecurity experts are urging companies to implement software updates to combat these risks. This comes as big tech companies like Microsoft, Amazon, and Alphabet…
- ELC Podcast - How Enterprises Actually Win with AI: Operationalizing Responsible AI, Engineering Guardrails, Trust Controls, and Systems Thinking at Scale — Sfelc
ELC Podcast - How Enterprises Actually Win with AI: Operationalizing Responsible AI, Engineering Guardrails, Trust Controls, and Systems Thinking at Scale The Engineering Leadership Podcast · Episode 255 # How Enterprises Actually Win with AI: Operationalizing Responsible AI, Engineering Guardrails, Trust Controls, and Systems Thinking at Scale ## with Murali Swaminathan Apr 21,…
- Inside Artemis' "AI vs AI" war | Shachar Hirshberg & Dan Shiebler (Co-founders, Artemis) — Shachar Hirshberg & Dan Shiebler (Co-founders, Artemis)
Inside Artemis' "AI vs AI" war | Shachar Hirshberg & Dan Shiebler (Co-founders, Artemis) # Inside Artemis' "AI vs AI" war | Shachar Hirshberg & Dan Shiebler (Co-founders, Artemis) In this episode of In Depth, First Round Partner Josh Kopelman sits down with Shachar Hirshberg and Dan Shiebler, co-founders of Artemis, the AI-native security platform that just emerged from stealth with $70M in…
Notable reactions
- Qwen3.6-27B Delivers Flagship Coding in 27B Dense Model — Let's Data Science
“# Qwen3.6-27B Delivers Flagship Coding in 27B Dense Model | Let's Data Science Published: 2026-04-22T17:00:31+00:00 Author: Let's Data Science ## Summary Qwen has released Qwen3.6-27B, an open-weight dense multimodal model with 27B parameters. The model outperforms its…”
- Octen Sets New Global Benchmark for Search Infrastructure; Launches World's Fastest Web Search API for the Agentic Era — Octen
“# Octen Sets New Global Benchmark for Search Infrastructure; Launches World's Fastest Web Search API for the Agentic Era Published: 2026-04-22T09:00:00-04:00 Author: Octen Type: Press Release ## Summary Octen, the search infrastructure company for the AI era, has launched the…”
- Building a Knowledge Base for Myself - by Lennon - Blunt — Lennon
“Building a Knowledge Base for Myself - by Lennon - Blunt # Blunt SubscribeSign in # Building a Knowledge Base for Myself ### Three weeks inside Andrej Karpathy's LLM-Wiki pattern as a designer Apr 22, 2026 2 4 Share I too, saw this [viral…”
- My agentic engineering workflow at an AI unicorn 🦄 — Jacob Bartlett
“My agentic engineering workflow at an AI unicorn 🦄 # Jacob’s Tech Tavern SubscribeSign in # My agentic engineering workflow at an AI unicorn 🦄 ### Shipping fast at the cutting edge Apr 21, 2026 ∙ Paid 3 Share Rate my setup. From left to right: Wispr mic, clawdbot box,…”
- Ben Lynn's Online Garbage: Notes on my notes on Git — Posted by
Ben Lynn
“Ben Lynn's Online Garbage: Notes on my notes on Git I was mildly shocked, but honoured, to find that the venerable (or at least popular!) Wikipedia references my tutorial in [its article on Git](http://en.wikipedia.org/wiki/Git_(software)). Interestingly, the entry states that…”