Boris Cherny

1.9K posts

Boris Cherny

Boris Cherny

@bcherny

Claude Code @anthropicai

San Francisco Katılım Haziran 2010
133 Takip Edilen447.9K Takipçiler
vas
vas@vasuman·
Thank you Claude Code for auto encrypting all communication over CLI. Very SOC2 of you.
vas tweet media
English
20
5
228
24.1K
Ric Wainwright
Ric Wainwright@RicWainwright1·
@bcherny I am finding Claude agents to be amazing but equally buggy - sometimes I’ll go back to the main screen and it just sticks. I have to close and reopen the terminal to carry on…
English
1
0
0
29
Boris Cherny
Boris Cherny@bcherny·
@MalikUDraz @vasuman That’s a bummer to hear. Is there something specific we could have done better? A specific feature or ux that broke?
English
0
0
1
36
Malik Draz
Malik Draz@MalikUDraz·
@bcherny @vasuman @bcherny I loved Claude months ago but literally cancelled my 20x max subscription and signed up for Codex. Please stop beta testing on us. No point of velocity if it’s making harness garbage.
English
1
0
0
28
Boris Cherny
Boris Cherny@bcherny·
@tsarchitect_ @AnthropicAI Claude does have self knowledge — does it not invoke the Guide agent? If so, mind running /feedback and sharing the id here?
English
5
1
13
3.1K
Leo
Leo@tsarchitect_·
hey @bcherny @AnthropicAI would be amazing if claude had something like an updating product-self-knowledge skill so it actually knows its own current capabilities. me and a few friends keep running into the same thing that claude often has no clue about what it can do or how its own features work. e.g. had to tell claude code to research about /goal to understand that it can do it
English
2
0
7
3.2K
darren
darren@darrenjr·
the anthropic team is still innovating the most in the harness layer and it’s not even close opencode, codex and the rest are just copying claude code features mostly sure claude code is incredibly buggy but you can’t deny they’re actually pushing the frontier
English
130
11
748
148.7K
chrisreedbates
chrisreedbates@chrisreedbates·
Hey Boris, maybe a stupid question here, but I looked for /bug in the terminal + inside VS code. No dice... I ran /debug, but it was post facto - so not sure how i'd reproduce the issue without replicating the entire session exactly from the last compression... Is there a different way to to get you the ID?
English
1
0
0
33
Boris Cherny
Boris Cherny@bcherny·
@numberkash Maybe. If you scroll down a bit more, we show which specific plugins/skills/mcps used the most tokens. I find that really helpful for debugging fast context usage
English
0
0
3
80
Boris Cherny retweetledi
Daniel San
Daniel San@dani_avila7·
Just found that if you scroll down in the Claude Code app on iPhone… Clawd starts jumping and walking around looking for apps to build
English
7
3
88
14.3K
Lotfullah Andishmand
Lotfullah Andishmand@lotfullah_·
Sometimes Claude acts like an experienced surgeon, doing the tasks in a few concrete steps. Sometimes it acts like a very polite person in a level that's annoying, it is asking permission for every minor tasks. @bcherny, let's give it a full access like codex.
English
6
0
80
29.7K
Noris Alsaadie
Noris Alsaadie@norisalsaadie·
@bcherny - thanks for the insight. How does Auto mode work exactly? I assume it uses a Judge LLM pattern where Claude is asked about the safety of the next execution (using a lower temperature setting)? I feel we need new tools in this space with a focus on security and sandboxing. There are so many frictions points right now...
English
1
0
1
148
iamnobody
iamnobody@jiangyan·
@bcherny @lotfullah_ /goal + dangerous (ofc, with design + impl plan ) can have claude code work for several hours while sleeping, can auto mode do that?
English
1
0
2
360
Boris Cherny
Boris Cherny@bcherny·
We use both together. In practice, for containers to be useful, you often have to punch some holes: GitHub, Anthropic API, kube, other MCPs. Auto mode makes interacting with these safer. It significantly reduces the risks of accidental data deletion, exfiltration, and prompt injection.
English
1
0
4
523
Yonatan Cale ⏹️
Yonatan Cale ⏹️@YonatanCale·
`claude agents` is a step up in productivity for me! Better than my iterm2 left-tabs. It still has some problems @ClaudeDevs >>
English
2
0
1
287
Boris Cherny
Boris Cherny@bcherny·
@jinthagerman Yeah we’ve been working on improving reliability. It should be feeling better lately, expect it to improve more
English
1
1
8
3.2K
Jin Budelmann
Jin Budelmann@jinthagerman·
@bcherny Classifier is down 90% of the time I try use it. It doesn’t do the thing I want it to do which is allow me to leave and trust safe actions still happen
English
1
0
2
3.6K
Boris Cherny
Boris Cherny@bcherny·
People often ask what my biggest tip is for getting the most out of Claude Code. These days my #1 tip is: use auto mode Auto mode means no more permission prompts. It is the key building block for multi-clauding: start a session, then while it runs, work on another session in parallel.
ClaudeDevs@ClaudeDevs

Two updates to auto mode: · Now available on the Pro plan · Sonnet 4.6 is now supported, alongside Opus 4.7 Shift+tab, and let Claude run.

English
377
260
4.8K
762.7K
NAV 🔋
NAV 🔋@nav7634·
@bcherny I've never been so sure that a reply was written by AI and I have zero beef with it! Love your work
English
1
0
1
136
NAV 🔋
NAV 🔋@nav7634·
@bcherny hey dude - me again! Would be amazing to be able to pick up on claude code sessions started by Routines. Use case: - support emails land in my inbox - Routine triggered - Claude notifies me with topline description via slack. - I need to dig deeper with claude code session but am AFK. I should just be able to log into claude app and pick up from the routine session but atm I can't because the session is burried in the Routine fucntion rather than appearing as a normal remote/cloud session. Many thanks!
English
1
0
0
197