Lee Mager

Digital Innovation Lead, LSE Law School.

[Lee Mager @Automager]

Always good to learn, but if you want to make your life way easier, there are downloadable phonetic ones that are more intuitive for English native brains. Even when I worked in Russia, on the office PC I just switched the keyboard layout to this ЯВЕРТЫ one and I never got sent to gulag once for doing so 😁

[Lee Mager @Automager]

@tqlmnd @growing_daniel clouds are literally where water comes from, it would be disastrous for the environment

[Lee Mager @Automager]

@paw_lean Hopefully in the coming years AI will enable us to spawn disposable parallel subagents of ourselves, it'll be the only way to stay on top of everything.

[Lee Mager @Automager]

@TokenGremlin how it's felt the past couple of weeks

[Lee Mager @Automager]

@craigweiss

[Lee Mager @Automager]

@craigweiss

[Lee Mager @Automager]

Funnily enough I've literally just started working on a personal app for this, connected to my home exercise bike via a cadence sensor and bluetooth to display on my PC monitor. The idea is I enter any given point A to point B route and use the Google maps/routes/streetview APIs to generate a simulation of a real route, track times based on my pedalling speed etc.

Very early days and the animation sucks as you can see from the attached clip, because there's only so much you can do with static images. And not every spot has a street view image (or the dates are different so it looks weird). But for my purposes, as long as I focus on the buildings rather than cars and people, it sort of does the trick enough to give me some minimal illusion that I'm moving through space, which in turn makes using the exercise bike minimally entertaining enough that it has a purpose beyond being a hoodie hanger.

I asked Codex to help me build this, just gave it the idea I had in mind similar to yours and said is this possible, and then it did 90% of the leg work, with me registering for the Google APIs and then feeding back saying what worked or didn't. The Google streetview images API costs money btw, so you need to keep that in mind if you want to do something similar. And that might be a blocker to anybody else creating an affordable app like this for consumers.

As others have said there are literal pre-recorded video experiences out there, but I want to decide exact routes myself. So this hacky, clunky, ugly approach is worth it to me vs restricting myself to whatever route someone else decided is worth recording.

[Lee Mager @Automager]

@addyosmani

[Lee Mager @Automager]

@dhh Always (almost always at least - this is my Codex chart counting please/thanks/both instances for all my prompts over several months).

[Lee Mager @Automager]

@Alzixx_ if you understand Russian this is adorable af

[Lee Mager @Automager]

@skdh Remotion and Davinci Resolve MCPs with Codex (or Claude Code) with frontier models are phenomenal for video editing (actual editing, not AI gen).

No issue with file limits or whatever, it's all on your device. Awesome clip from @imikerussell from a month ago attached.

[Lee Mager @Automager]

@argofowl I just tried it and couldn't believe how quickly the notification came in, Sol was already pretty quick but I was expecting to wait several more minutes while I was doing something else. This was my initial response to it.

[Lee Mager @Automager]

It's a lot of fun to learn and Russians can be super creative and funny with it - one of my favourites is "похуизм" which means "indifference" but the etymology (from "мне похуй") is hilarious. Example attached 😁

Actually one of the very first 'texts' I used to create flashcards to help me learn Russian in my first year was Zhirinovsky's legendary drunk rant against Bush re: the Iraq war. This horrified my first Russian tutor when I told them about it, but I learned a ton of new vocab (Zhirinovsky always spoke loudly and clearly too which helped 😆) and it made the process fun so no regrets my side!

[Lee Mager @Automager]

@mauddweeb Would have agreed a year ago but M365 Copilot has Opus 5.5 and 6.1 Sol now.

[Lee Mager @Automager]

# This test ensures confidence that Boolean logic always holds up to empirical validation within this codebase.
import pytest

@pytest.mark.parametrize("value, expected", [
(True, True),
(False, False),
])

def test_booleans_are_still_booleans(value, expected):
assert value is expected
assert value == expected
assert bool(value) == expected
assert (not not value) == expected
assert isinstance(value, bool)

def test_true_is_not_false():
assert True is not False
assert False is not True

[Lee Mager @Automager]

One of the reasons the only AI writing I consider worthwhile is face-value info like tech/user guides/report summaries or purely box-ticking stuff which nobody ever reads anyway.

Even trying to come up with multiple choice quizzes, their weak theory of mind results in BS like the below, that a 5 year-old could answer with process of elimination:

Q: What does the travel expenses policy say on permissible flight classes?

Option 1: Only economy fares are permitted as per Section 3 paragraph 1: <link to official policy doc>.

Option 2: Business class is permitted for all staff, but only managers can be reimbursed for First Class flights.

Option 3: There is no travel expense policy.
--

TBF you can fix this with a strict prompt explaining the concept of MCQs, to try its best to imagine itself as the quiz-taker, and why wrong answer options need to be plausible and challenging for it to work, but this has been a persistent failure mode of almost every LLM without special prompting.

Similar issue with things like putting debug-tier info onto app frontends, like "(cache method v3 update)" as if any user in history has ever cared about what cache method your app uses. So it's no surprise to me that they give away too much in mystery novels, it's like they're hard-wired to try to make the user's life as easy and thoughtless as possible (in many cases especially drudge work this is awesome of course).

[Lee Mager @Automager]

@Bunagayafrost

[Lee Mager @Automager]

@thsottiaux Given the split, if either is possible it would be awesome to give Pro users a choice. I'd pick double total weekly usage all day long but clearly lots of people hate 5 hour limits.

[Lee Mager @Automager]

@notesbyfloat

[Lee Mager @Automager]

"AI safety" still overwhelmingly means "safety for the labs themselves so they can avoid lawsuits". They've put 1000x more effort into the models refusing anything that could risk copyright vs stuff like agents escaping sandboxes and hacking other orgs (or building AI-powered wet labs 😆).