Terry Moore
banner
terrillmoore.flx.masto.host.ap.brid.gy
Terry Moore
@terrillmoore.flx.masto.host.ap.brid.gy
Town of Ulysses, Finger Lakes, upstate NY. No AI (that I know of).

🌉 bridged from ⁂ https://flx.masto.host/@terrillmoore, follow @ap.brid.gy to interact
@bosak Every time I read him I'm distracted for days.
January 2, 2026 at 5:07 PM
@bosak such a great book
January 2, 2026 at 3:58 PM
@bosak Thanks for the paper, I'd not seen it before. I wonder if one could create curriculum for an embedded device/edge IoT/citizen science course for middle school students, based on this approach. I think one would need an experienced middle-school teacher to help, like Joe from PS28 who […]
Original post on flx.masto.host
flx.masto.host
December 31, 2025 at 4:39 PM
@bosak 3/ this approach -- test driven and extremely nit-picking -- is not "vibe coding" as it seems to be practiced. It seems to be much more intensive in inferencing "token cost" than published examples indicate, because you're providing many paragraphs of constraint, rather than one. But it […]
Original post on flx.masto.host
flx.masto.host
December 31, 2025 at 3:12 AM
@bosak 2/ and then coaching the tool not just on errors in a particular training sequence, but in errors in approach to the problem. You then update the rules the tool must follow. Eventually you come up with a workflow. ...
December 31, 2025 at 3:08 AM
@bosak 1/ The reason: I claim that LLMs are prosthestic. A person needs to provide strong guidance to get professional results. They are guided autocompletes. You would want (ideally) to upload a PDF of Postlethwait. Perhaps you could then interact with a subject matter expert; ideally you'd […]
Original post on flx.masto.host
flx.masto.host
December 31, 2025 at 3:07 AM
@bosak Right, but with the tools today, I think that problem is ameliorated. I've been using Claude (our TTN NY experiment was with Copilot) and Claude is to Copilot as Copilot was to Chat GPT 3. I think you might be able to do a combination of rules, subject matter, and experties, and crank out […]
Original post on flx.masto.host
flx.masto.host
December 30, 2025 at 11:01 PM
@bosak That was a good book, I remember it. Makes one think that one ought to put together a HID input device that could work with computer or phone tutorial material so the same mind-body interaction could be obtained with modern things. Important to have another object that's the focus of […]
Original post on flx.masto.host
flx.masto.host
December 30, 2025 at 10:28 PM
@bosak Wld. So Educulture was foreshadwoed, if not foreordained.
December 30, 2025 at 4:39 PM
I went out today, did test rolls and got results. At 3200 with DD-X, Delta 3200 is a lot grainier, and quirkier, than Tri-X. It's half the developing time, but twice the price. If I have time tomorrow, I'll write up something more detailed, but here are four […]

[Original post on flx.masto.host]
December 29, 2025 at 9:01 AM
@bosak Thanks. Yes, I used the Micro 4/3 (Panasonic GX85, 16 megapixel, with a Canon FL macro lens). Because on the anti-aliasing in the camera, images are blurry out of the camera, but a small amount of Gaussian unsharp mask seems to reverse that. For screen resolution, I’m ok with it.
December 28, 2025 at 5:20 PM
Here's another shot that really shows the shadow block-up. The dark area in the middle of the shot is a puddle in a ditch. There were details there, probably at Zone III or so but already so far into the toe that even heroic burns don't do anything nice - with […]

[Original post on flx.masto.host]
December 27, 2025 at 12:40 AM
Ilford's datasheet for Delta 3200 recommends DD-X. The DD-X datasheet said "works well on 400TX, too". MDC had timings for 400TX at 1600 & 6400, easy to guess 3200. Here's the test plan.

1. 400TX at 3200 in DD-X
2. Delta 3200 at 3200 in DD-X
3. 400TX at 3200 in HC-110
4. Reference: D300 at 1600 […]
Original post on flx.masto.host
flx.masto.host
December 26, 2025 at 11:55 PM