Opportunity AI: 10 New Work Possibilities Unlocked by GPT-6 Astra

When GPT-6 Astra launched, many people quickly noticed it was meaningfully different from previous model iterations. In some areas it's remarkably powerful, yet in others the improvement is barely perceptible — or even feels like a step backward. The AI Daily Brief host described it as "both important and confusing" — a characterization that reveals something deeper about where AI development now stands: the value of new models is increasingly less about doing what you already do better, and more about unlocking entirely new capabilities you've never considered before.

The challenge is, if these are things you've never considered, how do you figure out what's actually worth doing with AI? That's the core question this article explores — a set of thought-starters around what we're calling Opportunity AI.

Efficiency AI vs. Opportunity AI

Before diving into specific use cases, we need to clarify one key distinction that runs throughout this piece: Efficiency AI versus Opportunity AI.

Efficiency AI helps you do your existing work better — faster, leaner, cheaper. It forms the foundation of how most people use AI, and is the primary source of AI's early value. To be clear, there's nothing wrong with Efficiency AI.

Opportunity AI unlocks entirely new possibilities. It's not about doing old things better — it's about doing things that were simply impossible before. The core argument here is this: companies that treat AI purely as an efficiency tool will miss the opportunities that eventual winners won't miss. The reason is that all efficiency-based use cases eventually become table stakes — resetting the entire industry's baseline expectations. Meanwhile, companies that genuinely bet on discovering new opportunities, even when those opportunities seem unrelated to their core business today, are far more likely to reinvent themselves for the new era.

What makes Astra interesting is that it's the first model that genuinely seems to dance between both.

It's worth noting that feedback from power users on X is mixed, particularly for coding. Francesco called it "possibly the smartest and dumbest model I've ever used," frustrated by the "ridiculous shortcuts it takes just to get things barely working." OpenCode's Dax mentioned that some team members rolled back to Sol, with costs doubling and hard to justify. But what's genuinely exciting is Astra's performance in entirely new capability areas — video editing, 3D design and modeling — which is precisely where Opportunity AI shines.

This framework isn't unique to AI. It borrows from the classic strategic management distinction between Sustaining Innovation and Disruptive Innovation — the former improves along existing trajectories, while the latter opens entirely new markets or use cases. Efficiency AI maps neatly onto sustaining innovation: same work, same processes, same business models, just with better cost and speed metrics. The problem is that when competitors gain access to the same tools at the same pace, efficiency advantages compress into industry benchmarks rather than lasting competitive moats. Opportunity AI operates on a different logic — closer to "leverage point" thinking. Find the ideas that have been shelved due to technical or resource barriers, and when those barriers suddenly drop, first movers capture a time-window advantage rather than just a cost advantage. This also explains why Opportunity AI may matter more to individuals and small teams than to well-resourced large organizations — the latter's organizational inertia often makes "exploring what's never been done" far harder than "doing existing things better."

Why Most People Face the "Blank Page Problem"

The honest admission here is that if you simply tell people to "go use AI to find new opportunities," the vast majority will hit a massive blank page. People don't carry around a complete inventory of everything they could possibly do — our imagination is shaped by our work, our tools, our experience, and what we see the people around us doing.

The automated system for slicing podcast episodes into shareable clips and pushing them into a social and video production pipeline wasn't built by sitting around waiting for the technology to be ready. It was stumbled upon through constant experimentation and problem-solving.

So a practical shortcut is: watch what people doing different work are doing that you think is cool. Many so-called "new opportunities" are actually things others can already do that you couldn't do before.

10+ Opportunity AI Thought-Starters

The following thought-starters are organized around Astra and Fable 5.1, each combining a high-level concept with concrete applications.

Make Marketing Something You Can "Play"

Historically, marketing content has been a combination of visuals, copy, and video. AI introduces a new dimension: interactivity. If Astra can create — or even recreate — games, why not invite users to explore a world, play a role, or complete a challenge? The rules of the game itself can embody your product argument. Imagine letting players personally experience how coordination costs rise as headcount grows — making them viscerally understand the core insight behind your consulting advice. In the past, marketing games were constrained by extreme scarcity of development resources. Today, a solo founder could try this in a weekend.

Build a Video Production Pipeline

The AI Daily Brief video clips aren't made with specialized editing tools like Opus — they're produced through a custom Claude-powered pipeline: ingest the script and raw footage, then run a suite of tools to produce the video. The prediction is that advanced models' democratization effect on video production may mirror what coding agents have done for software development.

The homework assignment: use AI to write a 60-second script, record it as simply as possible with an iPhone, then hand it to Codex or Claude Code to design visual motifs, transition elements, and layered graphics — and ask it to build a full production pipeline that outputs a finished video when you drop in the source footage.

The concept of a "pipeline" originates in software engineering and data processing — it refers to chaining multiple processing steps into an automated workflow so that data or content flows from input through each stage to produce an output without requiring human intervention at every node. In a video production context, a typical AI-driven pipeline might include: script text parsing → scene segmentation → asset matching → subtitle generation → graphic overlay → format export, with each step handled by different tools or models, connected by code glue (such as scripts written by Claude Code or Codex). The core value of this architecture is "build once, reuse indefinitely" — high upfront investment, but near-zero marginal cost. For creators without technical backgrounds, understanding the pipeline mindset matters more than mastering any specific tool: rather than thinking of AI as "an assistant that helps you edit video," think of it as "an engineer who helps you design and maintain an automated production line." The two mindsets produce very different usage patterns and unlock very different scales of value.

Explorable Product Demos and Shapeable Proposals

Traditional sales demos must commit to a fixed presentation order. Interactive demos can organize themselves around actions — open this, isolate that component, inspect the arrangement, review the results — letting customer questions determine the path.

Going further: customer-shapeable proposals. A proposal already contains an implicit model — assumptions about workload, parallelism, and available resources. When a client asks "could we bring in another department?" you mentally re-run that model. An interactive proposal makes that reasoning directly available to the client: "want it done faster" immediately translates into "if the team can engage more frequently we could finish sooner, or we'd need more capacity to run things in parallel." This one likely has both Efficiency AI and Opportunity AI value, and may well become standard practice.

Business Decision Simulators

AI's ability to absorb information and generate large numbers of possibilities adds a genuine dimension to strategic decision-making. Many disagreements actually contain multiple nested disagreements — "we need to hire another person" might mean the process is too slow, work distribution is uneven, or someone anticipates a future demand surge. Building a simulator forces you to make those assumptions explicit, making the consequences of "what if" scenarios auditable. There's a common pattern underlying these ideas: they're all essentially about building what-if machines.

3D, Customer Story Documentaries, and Practice Environments

Much of the early excitement around Astra came from its ability to work in 3D — building 3D walkthroughs of Zillow listings in Blender, or creating three-dimensional objects that significantly improve comprehension through rotation and interaction. The suggestion is to focus on asking: where would three dimensions be transformative?

You can also take existing assets — customer interviews, project retrospectives, screen recordings, screenshots — and use Fable 5.1 or Astra to assemble them into mini-documentaries, making the evidence concrete within a problem → invention → outcome narrative.

Another exciting direction is practice environments: simulate a situation before it happens in real life. Have employees practice handling difficult client conversations or tough exchanges with management, and receive immediate feedback on their judgment. This will likely catalyze an entire category of professional development experiences.

Turn Your Expertise Into a Product

The final thought-starter — and possibly the most relevant for knowledge workers — is taking a specific slice of your judgment and turning it into something others can operate themselves. Expert help usually takes the form of conversation, but inside that conversation a lot is happening: gathering context, recognizing patterns, noting exceptions, ruling out tempting-but-wrong options. The interactive web app described here is itself an example of productizing thinking about Opportunity AI — it can't match a one-on-one conversation, but it delivers a meaningful portion of the value to a much larger number of people. That's Opportunity AI. And in contexts where you're explaining the same thing to different people internally over and over, it's simultaneously Efficiency AI.

Converting tacit knowledge into reusable products is a longstanding challenge in knowledge management. Tacit knowledge refers to judgment developed through experience that is difficult to articulate — experts often "know" the answer but struggle to explain why they know. Traditional knowledge productization paths (books, courses, consulting frameworks) can only capture the parts that can be made explicit; vast amounts of contextual judgment and pattern recognition remain locked in the expert's mind, impossible to scale. AI changes this: through deep interaction with models (extensive Q&A, case training, prompt engineering), it becomes possible to "externalize" an expert's reasoning path into an operable interactive interface, giving users output quality that approaches expert-level advice on specific questions. This doesn't mean experts are replaced — it means the expert's time is liberated from one-on-one conversations and embedded into the product itself. For consultants, lawyers, doctors, coaches, and other judgment-intensive professions, this is a critical path for converting personal brand into a scalable asset.

Closing Thought: Let Yourself Be Stretched

Efficiency AI and Opportunity AI are not in opposition. The goal of Opportunity AI is simply to stretch us — to ask: what is there now that we would never have done before, but might actually be able to do today? The value of these thought-starters lies precisely in the fact that they don't require you to already know what the opportunity looks like. For practitioners facing a new generation of models and unsure where to start, rather than fixating on a model's "mood swings" in coding tasks, it's worth redirecting some of that energy toward the genuinely new possibilities it unlocks.