Literally no one at this 9 year olds birthday has heard of ai.

Infinite Red is the expert mobile consultancy that serious clients bring in when they need to get stuff done without losing quality, trust, or their minds.

It's so creepy in a 2001 Space Odyssey when HAL says " I'm sorry can't do that Dave" I get that often now from @AnthropicAI

Don't forget to get your Chain React tickets! The only US-based React Native conference. We're co-presenting with Expo this year and it's going to be rad. • Portland, OR • Workshops July 29 • Talks July 30-31 Buy Tickets → ti.to/chainreact/chainreact2… Retweets very welcome!

Last night I ran the Night Shift on a pretty important sweep: checking my test suite for problems. It found 96 issues in 158 tests. Pretty good work, to be honest. The problems I had it look for: - Reaching into internals to “cheat” - Shortcuts that reduce confidence - Overly-specific / brittle assertions - Tautologies (asserts values it just wrote) - Weak assertions … can’t really fail - Wrong layer (unit testing when the question is integration) - Stubs/mocks that don’t really test real conditions, or fixtures that don’t align with real code (I lean toward using e2e tests wherever possible) - Dead code - “Simulating” multiplayer instead of actually setting up multiple instances and testing for real - Doing redundant setup when one test scaffold would work for multiple related assertions (for test suite speed) I had it first build a catalog of tests, and then work through that with subagents, scoring / providing analysis. And then one by one, with a commit per improvement, it went through and improved each one, running the tests as it went. Well worth the token spend!

I was thinking the other day that AI doesn’t just make you faster at your job. It also makes mistakes faster, confidence cheaper, “looks good to me” wayyy too easy. So what's the real question? It’s whether your team has a system for surviving it.

can codex with computer use use claude code tui? because that means we can automate claude code without paying extra

Bruh

Debuting his first album a deep soulful smooth jazz reading of @kenwheeler ‘s greatest tweets of all time.

I am now on attempt #15 trying to get @google Flash 3.5 to make the entire @threejs frame render on screen. Not going well so far. It has access to google devtools, it's even taken screenshots that "look" like it's working fine. But it isn't. Fail

Typo or is @antigravity coming out with a $100 plan? 🧐

🤔 @LowLevelTweets legit? techspot.com/news/112410-sec…

I disagree harshly, “No Taxation on my Tokenization!!!”. As we continue to bring the cost of inference down we’ll be saddled with a tax the government now counts on forever and will never want to release that back to us. It will grow massively as a percentage.

Based on responses @jamonholmgren is either the “worst developer on earth” or “fine” 🤣

$62B. That's what trial lawyers make every year in the US. ~$31B comes from car crashes. Halve the crash rate and you erase $15B of trial lawyer income overnight. Self-driving cars are going to unclog the courts as well as save lives.

I still don’t understand why people want their tools to have “personality”. I want “robot that writes correct code as quickly and efficiently as possible”.

.@theo you do you king