Türker Yıldırım

llm

  • A skill that was too long to be read carefully

    Note2 min read

    A two-thousand-word document describing this application’s architecture, which produced worse output than no document. A long document produces confident output that follows some of it, which is worse…

  • Splitting by task rather than by subject

    Note2 min read

    Four skill documents organised by subject, replaced by six organised by task, which is the whole finding. A document loaded because a task started is read; a document…

  • An agent with a scope and a list of what it may not touch

    Note2 min read

    An agent definition, which is mostly a list of prohibitions. The last instruction is the one that matters and it is the one a general-purpose assistant gets wrong…

  • Reusable skills and agent definitions, and what they are actually for

    Deep dive18 min read

    turkerdev/ai-skills and turkerdev/ai-agents, and eighteen months of retyping the same instructions. Context is per-session and knowledge is not.

  • Two years in, and the measurement problem is unchanged

    Note2 min read

    Eighteen months of trying to say something specific about assisted coding, and the same conclusion as the first attempt. Two years and the honest position is that the…

  • Reviewing code I did not write

    Deep dive10 min read

    Review as the skill that matters, and a volume of code that needs it. A change that looks right, passes, and is subtly wrong.

  • What we do not let it near, written down

    Note2 min read

    A scope decision recorded as a rule, because a rule can be reviewed and a feeling cannot. The list is short and it is the shape that matters…

  • A year of an assistant in the editor, measured

    Deep dive14 min read

    A year of using a model-backed assistant, and an attempt to say something specific about it. Every claim about this is a demonstration or a complaint.

  • A team disagreement that data did not resolve

    Note2 min read

    Two people with the same evidence reaching opposite conclusions about a tool, which is a disagreement about values rather than about facts. A disagreement that survives agreement on…