Spoofer3

@marick@mstdn.social First example, I was updating my writing instructions and include: ... Precedence: These are surface-form bans. If a prohibited form is the only accurate way to express a technical point (e.g., a real "not X but Y" correction of a specific error), correctness overrides the ban — but the burden is on genuine informational need, not rhythm or emphasis. and got back an aside: Precedence clause has an exact ancestor — Orwell's six rules in "Politics and the English Language" end the same way: "Break any of these rules sooner than say anything outright barbarous." Working as instructed and kind of interesting.

@simulo a limited take, I think it depends on where you are in your journey. At the beginning, I appreciate someone else's experience and recommendations... after I've become more knowledgeable or expert myself, probably not as much, until I learn some more and are able to appreciate it differently.

@simulo my impression is that most engagement across social media is bots, for either 1) getting the 10 leads from a million engagements monetization strategy, or 2) pushing propaganda from some viewpoint, and that for mastodon, its just not bot friendly in the same way, thankfully, maybe due to the moderation and structural choices. There is still a real problem for finding a funding mechanism for services like quality news that isn't ad or data harvesting based.

@shadow53@floss.social https://mastodon.online/@jamesshore/116552564152018598 talks about maintenance costs

@marick@mstdn.social The thing most immediately that comes to mind is, POSIWID — "the purpose of a system is what it does" (Stafford Beer)

@marick@mstdn.social I'd take it that i'ts complicated and not quite resolved in your mind to your satisfaction yet...maybe for similar reasons we procastinate...it's a sign of "not quite done" more than anything else.

@marick@mstdn.social I think both approaches have strengths for different purposes...like if I'm a numbers person, the graph is really impactful in the magnitude of the loss over time and distance. If I'm an emotions person, the painting may speak a lot more though they are two point captures in contrast to the data over time of the chart. Different audiences/contexts/purposes will have different suitability.

@marick@mstdn.social The successful scientists/engineers that I've observed focus a lot of their attention to the uncertainty part of their models. The uncertainty often has shape or relationships/sensitivities to the data that are often insightful, and also allow for determining when "good enough" is achieved. One scientist told me that the detail and length of the description of the problem is probably proportionate to the accuracy of the model. An enlightening exercise was looking through the references for NIST standards and understanding how they determine a constant (like a half-life). Often, they are great for everyday use, but maybe not so much beyond a certain precision and one might not have as much confidence in that constant after going through the experience. E.g. it might be the average of 5 master's thesises results and there might be quite a bit of variety in the rigor exercised.

@marick@mstdn.social I guess I'd be quite surprised if the squishy sciences ignored "p-values" and the related mechanism, which I lumped into "uncertainty" here, though aware there are lots of closely related items, including variance, confidence levels, accuracy, precision, etc. Anything that makes repeated observations (counts, measurements, etc) or makes predictions, etc has to deal with this. Another way I sometimes look at these models is that a model is asking/answering a question that either has been answered before, or this is the first time it's being asked. If it's not the first time being answered, then new results have to be compared to previous times it was answered and there should be a metric for assessing "better." ... If there isn't, then I'd say that the scientist isn't completing their job. Though after reading your series, I'm also realizing there is a lot of jargon (e.g. induction) that I presume I know, but is clearly being used quite a bit differently by the community you are writing on...so it's likely we agree a lot more than it might seem.

@marick@mstdn.social I'm not sure the full definition of 'trails' as used in this context...but if it is metaphorically akin to weak depth first search, in the sense that there is a thought and we have a new thought, so we make a save point and start looking at the new thought and when we get to the bottom of that rabbithole, we revisit a previous checkpoint...somewhat like a conversation or narrative...then this is very nearly how I'm (I'd argue successfully) using Claude-Code now days. Claude-code keeps track of claude.md files at the user/project/session levels that contain instructions for the "agent" part of Claude. There is also memory.md files that contain instructions/facts at the user/project/session levels that are germane to that project context...there is also a roadmap/plan.md file that is basically the "trail" part where I start with a "goal", decompose the goal into subtasks, decompose those into stages/phases, etc...Often the process of working on a "task" results in modification to the .claude, memory.md, or the plan.md. When a (sub-)task is complete, work continues on the next task identified in that trail. (There is more to it for successful claude development, but that is the part relevant to trails.)

@marick@mstdn.social Attributed to Will Rogers: "When you give a lesson in meanness to a critter or a person, don't be surprised if they learn their lesson."

@marick@mstdn.social There are "natural system" laws (e.g. the one's humans can't influence/manipulate (yet)" and "social system" laws, that might be asinine, but nevertheless, may have life and death consequences for individuals or groups...being more "right" doesn't necessarily keep one more alive. In the way narratives wander, this reminds me of the free-will, conscience, etc discussions where one of my constructs is imagine an agent with no memory....the agent is presented with a "choose a random number between 1 and 10" problem? The agent might "freely" choose three and the random number generator may "freely" return 5. That agent may be involved in repeated trials and can never affect a future outcome, nor dispute the randomness of the RNG, because they have no memory...so the agent's reality is that the game is fair and they have free will, despite the next layer's out reality that the RNG is not random and the agent always returns the same number. Substitute in some level of information sophistication/reasoning ability for "memory" and it reinforces (to me) how inadequate humans are for actually understanding anything one layer removed from what is right in front of our eyes/nose/ears.

@marick@mstdn.social How are LLMs at calling out serendipitous connections? In my experience, when they are focused on a goal/task, the LLM stays focused on it and 5%-10% of the time latches onto a errant/bizzare take that needs correcting. However, if I query an LLM for related topics, they are amazing at finding related items....I've not tried including a "serendipity" prompt as part of my system prompt, but that is a very interesting idea. I will add it and report back in a couple weeks.

@futurebird@sauropods.win maybe you would enjoy https://darknetdiaries.com/episode/153/ on the bike index. https://bikeindex.org/about

@mfowler@toot.thoughtworks.com I enjoy the "Fragments" very much and appreciate that you take the time to record and share them. Thank you.