Jul 9, 2026 · 49m · startup-ideas
GPT 5.6 SOL IS HERE! How to use it.
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Greg Isenberg and Dan Shipper demonstrate how to leverage OpenAI Codex and GPT-5.6 as a unified personal operating system, automating knowledge work and live-building an agent-native SaaS application.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Greg holds 20.5% of the talking time here. How this is scored →
Greg as informed peer 3.7 Guest teaching 5.2 Guest disagreement 0.5 Greg pushing back 0.3
speaking balance: gold is Greg, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 12:25 Rejecting the SaaS-pocalypse narrative
Dan forcefully rejects the popular doom narrative around SaaS obsolescence, arguing that vibe coding cannot replace long-term vendor maintenance and edge-case handling.
Hardest push from Greg ▶ 14:43 Operating system rather than super appGreg interjects to push back on Dan's 'super app' categorization, insisting that the setup functions fundamentally as an operating system.
Biggest teaching moment ▶ 35:13 The Pirate vs. Architect builder paradigmDan educates Greg on modern AI collaboration dynamics, reframing the classic technical co-founder need into pairings between exploratory 'pirates' and polishing 'architects'.
Greg holds their own ▶ 27:40 Greg leverages App Store distribution historyGreg demonstrates his seasoned founder background, contrasting token cost economics and predicting agent platform app store dynamics based on his past mobile app launch data.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Greg as informed peer | Guest teaching | Guest disagreement | Greg pushing back | Why |
|---|---|---|---|---|---|---|
| Welcoming Dan Shipper to the Show | 1 | 5 | 1 | 0 | Greg introduces Dan and enthusiastically yields the floor, admitting he doesn't know what training models entails in this context. Dan sets the stage by contrasting Codex Desktop against Claude Code and explaining how non-engineers can now build ML pipelines. | |
| Live Demo: Automating Email and Feeds with Tend | 3 | 6 | 0 | 0 | Dan demos his Tend email workflow and company pulse feeds that compile learnings into revised prompts. Greg observes that the interface feels like an executive secretary system from the 1960s Mad Men era. | |
| Evaluating GPT-5.6 Accuracy and Multi-Thread Orchestration | 5 | 4 | 0 | 1 | Dan explains how GPT-5.6 enables multi-agent thread orchestration and web browsing. Greg validates this by sharing his own realization that referencing across threads makes Codex function like a true OS. | |
| In-App Browsers, Proof, and Why SaaS Survives | 5 | 5 | 1 | 2 | Dan demonstrates Proof and dismisses the SaaS-pocalypse narrative, explaining why companies will still pay vendors for maintenance over vibe-coded CRMs. Greg clarifies the framing, noting it's really an operating system rather than just a super app. | |
| Live Build: Designing the Turnaround Maintenance Badge | 3 | 5 | 0 | 0 | Dan initiates building a live badge tool called Turnaround to prove software maintenance using CEIDA and CE Brainstorm. Greg enthusiastically encourages the live build. | |
| Comparing Model Tiers and Answering Product Decision Forks | 4 | 5 | 1 | 0 | Dan compares 5.6 to Fable, categorizing Fable as S-tier for agent orchestration and 5.6 as a practical A-tier daily driver. Greg inquires about relative performance percentages and notes the value of models having an opinion on product decision forks. | |
| Compound Engineering, Autonomous Goals, and UI Design Checkpoints | 3 | 6 | 0 | 0 | Greg asks for clarification on LFG and long-term goals. Dan breaks down the compound engineering loop and explains Codex design checkpoints while letting a long-term goal execute. | |
| The Opportunity for Codex-Native SaaS and Agent App Stores | 6 | 3 | 0 | 1 | Dan pitches the commercial opportunity for Codex-native SaaS apps where users bring their own compute. Greg draws on his own app store distribution background to highlight how platform app stores could fuel rapid customer acquisition. | |
| Evaluating the Turnaround Prototype and Local Models | 4 | 5 | 1 | 0 | Dan reviews the generated Turnaround prototype, critiquing its visual styling and comparing frontier models to local fine-tuned models for copy editing. Greg asks how Dan specifically evaluates local models versus proprietary frontier APIs. | |
| The AI Era Builder Divide: Pirates versus Architects | 3 | 6 | 1 | 0 | Greg asks Dan how he takes vibe-coded projects from 70% to 100%. Dan introduces the dichotomy of AI pirates versus architects, explaining that technical co-founders are being replaced by complementary builder dispositions. | |
| Advanced Codex Architecture: Mailroom, Pulses, and Router Threads | 3 | 6 | 0 | 0 | Dan details his Mailroom routing system, Slack agent bridges, and background heartbeat pulses. Greg expresses a desire to have a single exportable package for Dan's entire environment. | |
| Getting Started with Codex: Chronicle and Mindset Shift | 4 | 6 | 1 | 0 | Dan breaks down Chronicle and record-and-replay features, emphasizing the managerial mindset shift required to manage systems rather than individual tasks. Greg validates the workflow with SOP frameworks before Dan warns listeners against FOMO-driven over-engineering. |
Statements from this episode (22)
Dan Shipper: Anthropic holds 'mandate of heaven' while OpenAI is underrated
“I feel like Anthropic has the, has a little bit of the mandate of heaven right now, and OpenAI is quite underrated.”
Shipper: GPT-5.6 with Codex is best model for knowledge work
“Codex with five, six, which I find to be the most usable, most powerful, fastest model for, specifically for knowledge work and for coding, but like specifically for knowledge work”
Shipper: Advanced AI models enable non-engineers to fine-tune models
“Five, six and fabled some extent if you can get around the guardrails is makes it easier, easy enough to make a machine learning pipeline and run it that, and that's everything from like grabbing the data to making, like making more synthetic data to like actu…”
Dan Shipper: Users can feed video into Codex to recreate apps
“But honestly, you can literally just feed this video into codex and just be like, make something like this and it will do that.”
Dan Shipper: GPT-5.6 Accurately Handles 90% of Routine Emails
“I think that five five was fine, but not quite good enough. Like, I would say generally the email responses here, so like this email draft, for example, is like, It's as if I wrote it, like it's, maybe there's something that I would slightly change, and I migh…”
Dan Shipper: People Underestimate Codex's Multi-Agent Orchestration
“I think that people are sleeping on how, Flexible and powerful of a tool this is. Because, okay, so it's got the image model in it, and then it also has the ability, like it has the ability to change other threads. So, like, I can have this thread message anot…”
Dan Shipper: 'SaaS-pocalypse' is dumb; 99% will prefer maintained software
“I think the SaaS-pocalypse is so dumb because Hey, yeah, I'm like vibe coding this inbox app, but 99% of the world is going to want to use something that's maintained by someone else. Yes, you can one shot everything, but the I think the hard rare thing is mov…”
Isenberg: In the AI era, users will pay for software maintenance
“Yeah, I think yes, you can build anything, but people will pay for maintenance. That's sort of like the insight is that, you know, is if this thing works, I'm fine to pay for it.”
Dan Shipper: Fable AI autonomously emptied his team's bug backlog
“Like with fable, I was we were like literally The, our bug backlog was emptied out. You could just set it off and it was emptied out as long as you're willing to spend the money.”
Dan Shipper: Fable is too slow and expensive for average users
“It's because it's so, it's also so slow and it's so expensive that it's really only meant for Essentially like people who are doing machine learning or are orchestrating like fleets of agents, which is just a skill that most people don't have yet.”
Dan Shipper uses Codex Goal Mode for 20-hour autonomous training runs
“I'm training a model to do copy edits. And this goal is four hours, it's been going for four hours, and I just like let it go, but four hours is honestly very short, like I, over the weekend I had one or two that went for like 20 hours each, and so it's just a…”
Shipper: Rates GPT-5.6 Design Skills a B, Up from C or D
“This is another nice thing about 5.6 that's pretty new is, it does this like design checkpoint. So it's really, they're trying to make it better at design, and I think it's not, Perfect. Like it always wants to do warm paper backgrounds for some reason. I don'…”
Shipper: Agent-native SaaS restores traditional software margins without token costs
“In this world, you're just a SaaS provider. You're just hosting software. So there's no token costs. And so your margins are back to being like regular, a regular software business instead of being an AI business, which I think is great.”
Shipper predicts OpenAI and Anthropic launch agent app stores within a year
“And I bet you within a year the opening eyes in the anthropics of the world will have app stores for apps that are built specifically to be used inside of these services. Maybe even cursor too. I don't have any insight on any of this.”
Shipper: Anthropic's Claude and Opus remain ahead in visual product design
“I would go to Claude for the actual design, and then port it back in here, because I still think that Claude has a better sense, especially Fable, RIP. But even, even four eight would be better at this. And I think for a product like this, being able to see it…”
Shipper: OpenAI and Anthropic do not really offer fine-tuning anymore
“In that case, like, you have to use open-source models, because none of the major model companies, or at least OpenAnthropic, don't really do fine-tuning anymore.”
Shipper: Frontier AI models cannot copy edit like human copy editors
“These models can't do copy edits. They can't do copy edits like a copy editor can. And it's weird because copy editing shouldn't be that hard. At least copy editing to like 70% or something like that.”
Shipper: AI-era builders divide into exploratory pirates and polishing architects
“I make this distinction between two different types of people in the AI age or two different types of builders between pirates and architects. So a pirate is like, I'm just going to get it to 70%. I'm just going to like explore the field of possibilities and b…”
Shipper: Founders no longer need technical co-founders because everyone can code
“And I think that's a new and interesting divide where before we were like, I need a technical co-founder and that's not really it anymore because everyone can code. I think it's about your set of sensitivities and desire to make something new and go fast versu…”
Dan Shipper: Codex Chronicle captures local screenshots for user context
“Like, one of the interesting things about Codex is it has this thing called Chronicle, which is it automatically takes screenshots of your computer and locally turns it into locally turns it into a, Feed that gives it context on who you are and what you're doi…”