Daniel Schemann

Here it is! As many requested, the Tutorial about Godot with Rooms and Doors. 👉🚪👈youtu.be/ErOobMFODqA

AI assistants sound convincing, even when they’re wrong. I'm training our own model for orboto. First internal test: 0 made-up answers where the data has no answer, and 61 % vs. 44 % accuracy against a ~7x larger model. Smaller. More precise. At a fraction of the compute.

Yeah, I'm also into music. Here are some VSTs I'm working on: PULSE - dance bus & sidechain compressor VELVET - dynamic resonance tamer & air exciter HORIZON - sub-mono & mid/side stereo TITAN - true-peak limiter with LUFS meter Do i miss something?

I Like Trains!

How it started / How it's going

Yesterday, Opus 5.5 created demo accounts for us and was testing our system. Today, it refused! I was curious, as he is now saying he can't do that, and I asked who decided this. 👇

That’s a feature I have implemented really fast when working with multiple agents on different systems. A coworker on one project can send my agent a request to build and publish an update without requiring me now (within set limitations and rules). We leave our systems running so that while we are away, the coordinator can use our system for building and testing. The first time it was creepy, but we got used to it that they talk with each other.

Promised the table. n=16 reviewed tickets, 202 agent runs, same fixed Fable reviewer, same rubric. The 0 vs 3 on invented claims did not survive: 1.3 per ticket for both Sonnet 5.5 and Opus 5.5. 2 of 16 first diffs were mergeable as is. Quality comes from the fixed review, not from the build model. What did survive is cost, and the driver is not the model. Cache reads are 96 to 99 percent of the input side. Opus build: 171 tool calls, 263k context per call, 22.92 USD at list price, 78 percent of it cache reads. Sonnet build: 45 calls, 121k per call, 1.91 USD, 55 percent cache reads. Different ticket classes by policy (Opus got the multi-surface and security work), so a signal on the levers, not a verdict on the model. The lever is calls times context per call. Correction to round 1: my token numbers came from the agent summary, which omits cache reads. About 10x low.

Just noticed that I was opted-in to share my Unity editor data when I never did that... You know Unity, you have to ask people and let them opt-in and not just pre-fill it!

We A/B tested Claude Sonnet 5.5 against Opus 5.5 as coding agents on 3 real tickets: same prompt, same code base, one fixed reviewer with one rubric. Contained tickets: Sonnet wins. Fewer findings, zero invented claims, faster, half the price. Multi-surface work: Opus delivers the more complete slice, but overstates in every report. Our rule now: Sonnet by default, Opus for migrations, security surfaces and cross-cutting stories. Always a fixed reviewer, never the model that built the diff.