@Teacherman1986@TID012312 Think: what if you tie a small rope to a bat handle and spin it around at high speeds and hit a pitched baseball, is it going to go far?
What if instead you attach it to a robotic metal pivot arm with bolts with a hydraulic system with half the bat speed?
Which hits harder?
@Teacherman1986@TID012312 EV is F=mass*acceleration.
Acceleration means how efficiently the hitter transfers force into the ball.
Chasing bat speed with long pushy swings puts the arms in position to ABSORB energy instead of transferring it.
HLP Squaring is a stronger position even if slower bat speed.
@Teacherman1986 Powerful because big guy, big stretch comes from big forward movement. Average is low because forward movement makes him extremely timing dependent
ONCE is back! It's now a full-fledged application server for running dockerized web apps, like Campfire/Writebook/Fizzy or your own vibe-coded adventures. Zero-downtime upgrades, scheduled backups, and a gorgeous TUI with hyperdrive graphics. Enjoy! https://t.co/WaLSMms2fr
@RunNGun8 Modern teaching has the arms work independently and with extension, and as a result even if bat speed is faster it doesn't translate into better EV because force is absorbed by the arms during contact
@RunNGun8 The position the body is in when contact happens is so much more efficient at transmission of force into the baseball, because the arms are not fully extended, they are reinforced by the entire muscle chain.
Windows Notepad.exe now has a remote code execution vulnerability.
You read that right.
Notepad.exe, which used to be a simple text editor, has had so many network connect features added (including AI and Microsoft account subscriptions)… that it now has security vulnerabilities.
This CVE is rated as “SEVERE” and given an 8.8 score.
https://t.co/0FMfoho7f9
I went from entirely hand-coding to getting high volumes of useful AI generated code in less than a month, with only extremely rare and manageable hallucinations, though just one change to my programming practice. Which I will now describe.
1. Write and maintain a context file
Before you write a line of code, dump your thoughts about what you want to do into a text file. As you have more thoughts about it, add them.
It should be designer's notes. It could be a to-do list. It should be where you sketch things like file formats and protocols and module organization.
Eventually, pieces of this file will migrate out of it and become formal documentation. But it is not formal documentation. It is you dumping your thoughts to where the LLM can read them. Even your half-assed, speculative thoughts.
The process of writing and updating your context file will help you achieve clarity about your design. And of course, it is a huge, useful prompt.
I've always had a tendency to include designer's notes with my source distributions. It was a natural evolution from that to having a context file from the beginning of the project. The only conceptual breakthrough was one I realized that it was okay for part of the document to be a graffiti wall.
2. Design from the middle out.
Your program has an engine. Design that engine as though it were intended to be a reusable component, even if it isn't. That is, specify the inputs, the outputs, and the invariants.
The thing I'm saying to avoid is classical "top-down" design, which tends to over-constrain the visible interface of your program. Also, absolutely do not worry about the low-level details yet. Don't sweat performance, in particular.
Once you have an engine that works and you can unit-test, then you can build and tune the rest of your program. Be prepared to throw away the interface scaffolding you built to test the engine (but keep the unit tests, of course).
Why you do things this way: it's how you get proper internal modularity and separation of concerns - and a result that's maintainable and understandable. An LLM will not usually give you these things if you just tell it what you want the entire program to do.
Your engine may have a bunch of parts that you need to design-sketch and specify separately. That's okay; the point is, you're carving the problem apart in your head into pieces that you can specify crisply to the LLM.
Just keep biting off pieces of the problem that you can specify well until you get to the point where you can start plugging them together into production code.
An example of this kind of piece is: a parser for some sort of input format. That piece is done when you can see your program dump a digested syntax tree that has the shape you had in mind.
Among other benefits, partitioning the work like this means that individual design sessions are less likely to overrun the context limit of your LLM and start getting crazy behavior.
This isn't a change in my practice, because it's the way I've been doing things since I was a fledgling programmer.
Aaaaand...that's it. That's all there is to it.
Excited to introduce: Gamma 3.0
A generational leap for the world's most popular AI presentation tool.
Two major changes:
1. Gamma Agent - with one prompt, you can make sweeping edits across the presentation.
a. Say 'make it more visual' and it will scan each slide for data that could be visualized better.
b. Upload a screenshot of rough notes you want to turn into a slide and it will synthesize the info, search the web to fill gaps, and make a table for you to match your brand or theme.
2. Gamma API
Say you have a sales deck you want to reuse as a template. You could build a Zapier workflow to convert all of your meeting transcripts into personalized decks, automatically sent after every customer meeting.
Bonus:
We’re also launching Team & Business plans, bringing visual storytelling to organizations of all sizes, and a new Ultra plan for our power users looking to push the boundaries even further.
PowerPoint has over a billion users, but the vast majority struggle to use it effectively. Why is that?
Because making PPTs is still slow and tedious.
We want Gamma to feel different.
Our goal is to both lower the floor and raise the ceiling, so that anyone can get their ideas out there.
Can’t wait to see what you make with 3.0.
I don’t use AI to write code.
I use it to sharpen how I write code.
ChatGPT-5 = prompt engine.
Claude Code = refactor machine.
Me = 15 years of product engineering, taste, intuition.
The tools explore.
Experience decides.
That’s the game.