OpenAI released GPT-6 Astra this week, describing it as its most capable model for complex, end-to-end work. According head of product:
And the usual X mafia went crazy about “Astra is AGI” posts, I also have to jump on the bandwagon and see it for myself.
My refactoring project: AI Hedge Fund Analyst
I’ve been building an AI stock research system that continuously researches companies and markets and generates investment analysis. It started with a wiki to help me remember every AI bottleneck stock I was reading about. It will scan X, Reddit, news sites and various sources regularly, and compound knowledge to the wiki. Then I can easily recall them:
And now it has evolved into a self-learning, decision system and dashboard. My favorite is actually the daily market brief and its “conviction list”.
Two goals I wanted to accomplish during the refactor:
Untangle the wiki, automation, and dashboard code.
Goal: Create a repo for the dashboard so I can tweak look & feel without changing underlying data.
Redesign the dashboard for easier reading and create a design system
Goal: Consistent look & feel when new data is published
I cranked it up to High Effort and gave the agent the goals
I used to go back-and-forth with my agent to fix things piecemeal. With this refactor, I used /goal so I didn’t have to babysit the process. Tibo also gave us two weekly resets to play with Astra, so I cranked it up to High Effort.
After meta-prompting and brainstorming with AI on the goal, it gave me a .md file with specific design system, color schemes, layouts, and how to present text information, including tickers and green/red indicators. The spec defined:
what should change
what should be preserved
how the result should be verified
what counted as finished
where human approval was required
pro-tip: always ask AI what else you didn’t think of
The results were disappointing
After working for 3.5 hours of working on High effort, I eagerly awaited my brand new dashboard.
Maybe because UI isn’t Astra’s strong suite, or my goals too specific (or not specific enough), or the fact I should have been using Codex instead of Hermes as harness, but the results require another few hours of tweaking. My first question was - what happened to my Market Brief (my favorite feature of the dashboard).
After another few hours of tweaking and using three weekly resets, it finally came together. The entire specification → implementation → review → revision → production verification window was roughly 12 hours and 40 minutes elapsed, including my reviews, interruptions, and analytics setup.
A separate source-control and portability goal took another hour enforcing the design system.
Total time to refactor my UI: 14+ hours
Before vs After
This feels different from vibe coding
Similar to Fable 5.1, Astra is designed for long-running tasks. Instead of telling Michael Jordan how to shoot a basketball, you should tell the model more like “I want to win this game, figure it out.”
Out of the 13+ hours I spent with AI, I probably only actively worked on this for a few hours. Other times I spent driving kids around and playing Civ 7 (if you’re curious, it’s my first time playing Civ, and learning curve is too steep).
It handled hours of implementation, debugging, testing, and verification while I moved up a level and spent more of my time defining outcomes, reviewing the product, finding failure modes, and deciding what was actually good enough to ship.
Tokenmaxxing is also at another level. I have never hit my Pro plan limit before, but Astra burned through my weekly usage limit three times. Thankfully, OpenAI handed out free resets to try Astra - great business idea to unlock builders
.
Pro Tips
My take - don’t use Astra unless you have a lot of money to spend, a lot of patience to wait around, or solving gnarly complex problems. Everyone’s tempted to use the latest and greatest, but in this case, the latest isn’t the greatest at addressing most use cases.
Tip #1: Always use Codex, which is specifically tuned for Astra.
Tip #2: /goal command can work wonders, especially for long-running “AGI” models like Astra, but make sure you have the budget and very, very clear goals. Otherwise you’ll spend more tokens fixing the results.
Tip #3: Use Astra Low effort for everything.
Tip for long-running jobs:
It’s live
The result is AI Hedge Fund Analyst v0.1 - have a look.
You can browse the Daily Brief, company research, valuation models, and top stock picks. It’s a public research tool and an experimentation right now, not investment advice.
For the nerds, here’s dashboard repo. Note: actual data and automations are the backend that powers this agent, and isn’t available in repo yet. I will slowly add them.
But I’m actually more curious about something else:
What improvements should I make so you’d pay for this?