Sabitlenmiş Tweet
Cory House
41.7K posts
Cory House
@housecor
I help dev teams be insanely productive with AI. Courses: https://t.co/D5emROQa4J & https://t.co/6L1fD898mh Consulting: https://t.co/Qfp4Tfp3jf ⚛️
Kansas City Katılım Ocak 2009
806 Takip Edilen153.4K Takipçiler
Created this new project a week ago and already the number of files in the project root is dizzying.
English
@timneutkens Thanks Tim! I've applied it. Totally forgot about this, so thanks for the reminder.
English
Next's messy caching API history makes working in AI difficult.
Today Claude confidently said "fetch calls are cached by default."
Nope. They're not.
They were in 14, but thankfully aren't cached by default in 15+.
English
Oh yeah! Forgot Next offers a slick compressed docs index for your LLM's instructions file.
See here: #try-it-yourself" target="_blank" rel="nofollow noopener">vercel.com/blog/agents-md…
This assures it uses all the latest docs.
Thanks @hjoelxyz
Cory House@housecor
Next's messy caching API history makes working in AI difficult. Today Claude confidently said "fetch calls are cached by default." Nope. They're not. They were in 14, but thankfully aren't cached by default in 15+.
English
Sometimes using AI feels like being a tech lead.
I specify the goal, and let the AI handle the details.
English
It’s amazing how much time you can save by not arguing with people
If someone isn’t engaging in a good faith discussion I just move on
English
I’ve been waiting for a built-in OpenClaw alternative. It’s here.
Felix Rieseberg@felixrieseberg
We're shipping a new feature in Claude Cowork as a research preview that I'm excited about: Dispatch! One persistent conversation with Claude that runs on your computer. Message it from your phone. Come back to finished work. To try it out, download Claude Desktop, then pair your phone.
English
@housecor @JimTesorero Now I know you are lying. No states have 3 cents per kwhr.
English
I don’t thunk enough people realize how much cheaper an EV can be to operate.
Today I took a 200 mile trip.
My gas SUV gets 20 mpg highway. So I’d have spent $35 on gas.
Instead, I took my EV. It used just $2.40 in electricity.
Over 14x cheaper!
And the EV is smoother, quieter, easier to maintain, and MUCH quicker.
English
AI code review: That feature doesn't exist.
Me: I've been using it for months.
GIF
English
Poll: Do you typically give your LLM full permissions?
Examples:
Claude: dangerously-skip-permissions
Codex: danger-full-access
Copilot: allow-all-tools
English
This is pretty impressive.
Using Claude Code with Opus 4.6 - It's configured to run my tests after each change to validate its work.
The tests failed, so Claude stashed its work and confirmed the test failure was pre-existing - it was occurring before Claude's change.
Slick.
English
Hour long convo with my friend @JTowerMI about coding effectively with AI, and the key decisions that are easy to overlook:
Trailhead Technology Partners@TrailheadTech
In this episode, host @JTowerMI sits down with @housecor, author, consultant, founder of ReactJSConsulting, and a 12-time Microsoft MVP, to explore what it actually looks like to code effectively with AI today. hubs.ly/Q046Ym0T0
English
@cyrilgupta I help people use it but that doesn’t mean it’s always the right tool.
English
@housecor But I'm seeing you help people be insanely productive with AI. I'm not getting this tweet at this point
English
Even in the age of AI, I can do many tasks more quickly manually than via prompting.
Entering the right prompt, granting access, waiting for response, reading response, reviewing results, and iterating on results is often more time-consuming than just manually making a change.
English
Delegation adds overhead.
To delegate to a human, I must clearly convey the task, answer their questions, validate their results, and ask for tweaks after reviewing their results.
Same story with AI.
Takeaway: With both humans and AI, sometimes it’s quicker to do it myself.
English
Even as an AI optimist I gotta admit this rings true
AI agents work best for mid-sized tasks.
Sean@shonson_dev
@housecor Exactly. It's a tough trade off. Using AI for small changes is a waste of time, using AI for big changes is often a disaster.
English
Rule: If the prompt is likely to be longer than the code change, I should probably code it myself instead of using AI.
English
@kyzookyzoo Yes though if it’s shallow tasks might be faster to do one at a time
English
@housecor Then it depends on what you have on your to-do list
Sometimes it is a lot of shallow tasks that needs multitasking
Other times it is deep work that needs singular focus
English
Multi-tasking doesn’t necessarily make me faster, and it increases my task switching overhead
kyzooghost@kyzookyzoo
@housecor Yes if you focus on a task you could do it faster, but that's the thing - you can't multiply your focus, you can multiply AI agents
English
@liran_tal I see lots of examples.
Change name
Fix spelling
Remove unused function
Change some copy
Add an HTML attribute to markup I’m viewing
List goes on
English
The trap with AI is it’s hard to predict when it would be faster to do myself.
Steve Oak@ForbiddenSteve
@housecor i spent forty minutes trying to get a prompt to format a table yesterday when i could have typed the whole thing out in five, so i am clearly the problem here
English