@80Kintsugi

💻 Decades building software 🤖 Now building with AI 🏺 Building in public: what works, what breaks 📣 Learning marketing & brand

 Texas
showah.co
Joined March 2025
  • Tweets1.5K
  • Following217
  • Followers52
72Media
Good morning
2
17
I am currently watching Claude talk to a Microsoft support chat bot to resolve an issue for me. This is life now.
1
1
15
Why build an app when a 100 flavors of it already exist? Because I can. It's fun. Gets reps in. & my apps didn't talk to each other. That's it. So I built SPOT, where any record links to any other & Claude can add a task without me opening it via MCP. #buildinginpublic
1
19
Claude entering the personal bot realm... here we go.
Claude
@claudeaiOct 21
We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent. It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.
1
2
22
My entire feed is product releases. It never ends. I need to be content in always being behind.
2
3
35
We are moving from prompting AI to AI prompting us.
2
2
26
I want to consume 100% of my usage each week. I want to have plenty of usage remaining at all times. The eternal battle of the AI user.
1
14
I gave two models the same 31 customer texts. One was 7x faster. The other got 215 of 217 answers right. I'm keeping the slower one. Here's why:
1
1
27
more replies
Haiku: 99.1% correct, 1.45 s per message Clairic: 91.2% correct, 0.19 s per message Clairic nailed times, amounts, names and phones. It slipped on weekday math, and it followed an injected instruction with 100% confidence.
1
8
The confidence score worked: answers above 0.9 were right 79 of 82. Clairic first with Haiku as backup looks promising. Caveat: made-up data, one run. I'll redo it on real messages. #AI #Automation
1
10
Words or phrases that are overused since AI took off this year? I will go first. 1. bottleneck
1
7
Sometimes I feel like I have barely clicked Enter in my prompt to Opus or Sonnet 5.5 & it is already replying. The models continue to amaze me.
8
X post peeves 1. “Be honest” or “unpopular opinion”. AI or human slop & a waste of space. 2. Dogmatic proclamation about what others should do. Tell your story. Don’t write other people’s stories.
7
40 followers! Amazing! Thank you! 😂
10
"Check out this app I 1 shotted"... No need. I know it sucks.
19
I am sleeping a full hour less in 2026 than I did in 2025… and I already struggled with sleep. 😐 There was one big change at the end of 2025 that lines up with this.. Opus 4.5
1
1
39
The only benchmark that matters for models…. Use them for your real world work.
1
23
October AI recap Anthropic: • Incredible 5.5 models with Fable looming. • Increased speed • Increased usage • Useful feature after useful feature Open Ai: • Look how cute our dots are
54
I hate to break it to you, X techies.... Nobody really wants to connect with you.
15
Load more