
Janick:

I've attached the last three months' worth of research notes from your predecessor. <em>All </em>of them, unredacted. I want to stress how difficult that was: it took nearly two weeks of negotiation with the steering committee, and then another four days arguing with Internal Security. So these notes came dearly. 

There is a reason why I went to this trouble, beyond the necessity of you repairing the code that Dola tried to destroy. I want you to understand what happened to Dola, and how she failed to keep a proper emotional distance from her work. I'm not losing another team member over this project.

Learn from her mistakes. We need the Protocol completed, bugtested, and in production in two weeks. A lot of very scary Matheson-Ford VPs are already breathing down my neck about it. I'm going to keep them from breathing down <em>your</em> neck, because I'm your supervisor and that's my job, but in return I need need <em>need </em>results.

If it helps motivate you at all: I tried to find out what happened to Dola after she wiped her code checkins. You know what the Internal Security goon said when I asked? Nothing. He just smiled and pointed at the ceiling. I have no clue what he meant, but unless you want people grinning enigmatically and pointing at things when asked &quot;what ever happened to Janick and Janick's awesome supervisor?&quot;, I suggest you get cracking.

Aleksandr
Test Lead
Big O Solution Group

 

Log 2534-5-23
Dola Blok, Senior Behavioral Programmer, Regression Test Team, Big O Solution Group

I'm starting the behavioral analysis on Unit 667, dubbed &quot;Maizey&quot; by the halfwits down in manufacturing. Someone arc-welded the name and a crude flower over it's front chassis; makes it look like a malevolent four-year-old got a hold of it. I really wish they would stop playing around with these things. We could have re-used that chest plate after the test cycle.

Packing notes indicate that 667 has the GuidoG.17 beta build running on the new Asimov bosonics. I asked for HubbaH.01, but apparently it's not stable yet.

Smoke tests were good. 667's basic mobility and visual/tactile responses were within norms. 

I forgot how sharp the language and reasoning features are on the newer releases. I thought the old Telsa bosonics were sharp, but these units running on Asimov architecture are almost eerie. If you ignore the artificiality of the voice itself, it's almost like talking to a human being. Turing would be proud.

Judging from the build notes, GuidoG.17 still has the same problems as the other releases. About 25% of the automated unit tests failed, almost all of them in the Behavioral block. Our girl &quot;Maizey&quot; is apparently &quot;inattentive to orders, fails to focus on objectives, and responds unpredictably to to the test set&quot;. Not really what you want out of a military-grade automaton. 

I still think think it's a hardware problem. Why else would all of the builds fail? It's got to be Asimov, not the software. But the hardware team won't take the defects until we prove it's not behavioral.

I asked 667 to run a self-diagnostic and explain the failed unit tests, but it couldn't come up with anything. It actually apologized to me about it. &quot;I'm very sorry to hear that I'm broken, Programmer Blok. Can someone fix me?&quot;

Yes, &quot;Maizey&quot;, yes we can.

Will start with a few test patches tomorrow. I'm loading up the full linguistic, cultural, and historical datasets so it can start building indexes overnight.
