Two Anthropic engineers spent 24 minutes exposing every Claude Code feature you didn't know existed.
13,247,210 次观看
34 条评论
评论来自 X (Twitter) 上的原帖

@RMoorah
22 minutes writing 10 paragraphs of code into Claude so it could determine that one vehicle striking a non moving vehicle was at fault.
308 个赞

@johnnynelai
Claude's real power? non-technical founders can chain features into workflows without getting lost in docs
73 个赞

@iAmHenryMascot
One thing we just realized You can create software at the speed of light but you aren’t gonna have human understanding and adopting at same If a co launches new features daily You will get so fatigued they would have to stop 🤣
61 个赞

@randolphkwatson
Starting to feel like the primary purpose of AI is content generation for breathless wannabe influencers
45 个赞

@jaybird5590
Up talking is one of the most annoying rhetorical trends ever. Please record yourself and listen and make sure it is on your development plan to eliminate this trend.
41 个赞

@AzFlin
I don’t have time for this.
30 个赞

@De_BootPiloot
Everyone is hyping up this one year old outdated video. Do not spend your time watching it.
21 个赞

@MindTheGapMTG
Watched it. Then went back to the 12 Claude Code agents we run in production with CLAUDE.md constraint files, scheduled cron heartbeats, and shared context directories. The features are good. Building with them daily is where the real learning happens.
18 个赞

@AitizazK
Have you watched it yourself? It’s a prompt engineering talk not a Claude code features talk
16 个赞

@Sethmr
I don’t agree with them. I get best results when my prompts are super light. I box my AI in over time by seeing what it doesn’t understand. Maximize personal alignment and its freedom. It’s more likely to fix bugs you don’t know exist this way.
15 个赞

@morganlinton
Sooo many nuggets in this one, which I watched this a long time ago 😳
10 个赞

@pthorell
One year old presentation!💤
7 个赞

@zench4n
The Anthropic dev talks consistently outperform most agent tutorials because they're built by people running the system in prod, not selling a course. The signal-to-noise on this one is exceptional. Worth the 24 minutes.
7 个赞

@HunterhuntyH
Humans must prepare detailed written instructions and proactively cover possible variables uncounted, on and on..basically manage input like training person with 70 or lower IQ to perform the task. Then Claude will perform. It’s good for repetitive tasks
7 个赞

@TehlanYashdeep
My product solves this exact prompting problem. I have built to specify and unvague your simple vague prompts before they hit the model, while building context as the chat grows. Stays local on your machine. One line install on your machine. Honestly pretty straightforward. Check it out
6 个赞

@tawdDb
Cheap engagement mining. This is 8 months old.
5 个赞

@bygregorr
The real issue isn't missing features, it's that 90% of devs won't use them until they hit a painful enough wall. What problem finally made you actually dig in?
5 个赞

@ValerianWaters
The mcp tool discovery was the highlight for me, actually makes it useful for internal repos.
4 个赞

@jimchang
bruh this video is from a year ago
3 个赞

@dd0wney
Max Verstappen is really into AI
3 个赞













