Two-shotted a personality matrix analysis and life path simulation with opus 4.6.
It understands your skills/personality then presents simulated life paths. You then get a prediction of what would happen based on your life decisions.
Going to be sharing posts comparing gpt 5.2 and opus 4.5 for code review adding a new db migration. I'm on CC max and codex plus.
Scenario: I needed to change a field from null to required.
* gpt 5.2 caught issues with null/empty frontend states
* gpt 5.2 caught potential future issues sharing a database env variables for tests and non-tests that could result in wiping the prod db
@beyang I find 30-40 minute loops to be great - focused enough that I can still review the output but enough autonomy to get substantive work done.
idk how people are setting ralphs loose overnight
@alvinsng@FactoryAI been using the same to perform code review and it's been a huge help.
how do you ensure that all the rules are enforced to human standards? i find that oftentimes, even with lint rules, the agent can miss things. still better than none though!
Building the ability for a single claude code with ralph to chew through workstreams and tickets.
Lots of edge cases - graph traversal, ensuring the code quality stays high, retries, blocked tickets that require human intervention, etc.
Slowly chipping away!
Incredibly surprised how people manage claude code ralph wiggum on massive features.
1 hour: 44 files changed, 8000 lines added. new context per ticket with ability to access all previously completed ticket summaries.