Ben Holmes

13.9K posts

Ben Holmes banner
Ben Holmes

Ben Holmes

@BHolmesDev

The Whiteboard Guy @warpdotdev

Ordering more markers Katılım Eylül 2018
947 Takip Edilen39.6K Takipçiler
Sabitlenmiş Tweet
Ben Holmes
Ben Holmes@BHolmesDev·
A lot of developers are using agents to write their code these days. I was a hold-out... but Claude Opus 4.5 finally cracked me. So, I made a video walking through how I write code with agents end-to-end. We'll talk plans, rules, and everything else in my setup!
English
27
27
453
56.3K
Shashi 🇬🇧🇺🇸
Shashi 🇬🇧🇺🇸@Shashikant86·
Petra, Warp is cool apart from memory part (300-500 MB) which is ok for what I get out it.. Project/File Explorer, Code Review, Edit code if needed. Only thing I don't use is warp harness, I launch Codex/Amp/Claude Code from the warp terminal. TBH, I never logged in to warp and never felt I should.. (I tried and struggled a lot in the past so I gave up) .. Overall I love the product and it's amazing.
English
0
0
0
39
Shashi 🇬🇧🇺🇸
Shashi 🇬🇧🇺🇸@Shashikant86·
After trying multiple terminals iTerm2, terminal, Ghostty etc I settled on @warpdotdev, It takes bit of memory but does amazing job. Just gave a try to cmux, (Ghostty fork) looks promising but too glitch and not ready for me. Back t warp!
English
2
0
2
291
Ben Holmes
Ben Holmes@BHolmesDev·
@OseUghu Curious what's getting annoying. Anything specific?
English
0
0
0
44
zinj.
zinj.@OseUghu·
switched permanently to ghostty. warp was becoming really annoying
English
1
0
0
87
Ben Holmes
Ben Holmes@BHolmesDev·
- Curious if the SSH session is closing unexpectedly in a way that Warp could fix? Any repro information helps, including the command or SSH command you're running - We're working on vertical tabs - Say more on on-demand. We do have auto-detection for Warp's built-in agent - Is this when using an agent or just running commands?
English
0
0
0
110
Bruno Skvorc
Bruno Skvorc@swader·
Kind of falling out of love with @warpdotdev, wondering what people's other favorite terminal GUIs are. Used iTerm a lot, but realizing I would love something with vertical tabs that keeps context well, so that I don't have to search through tabs every time. Examples: - I have an SSH session in prod, but it'll close every now and then. The tab will lose history, so ⬆️ will not summon last command, but the last command of the user's history across any tab, and this is destructive to flow - I have a tab for each context I push from, as I push regularly from different repos. I frequently lose sight of them because of horizontal tabs, and the coloring / renaming in Warp isn't very good imo, it still makes me search for things too long - on-demand AI integration would be nice, but ideally faster than the one that Warp uses, and I would not want to run a full codex/claude session. Fewer mini-frictions. - Warp has a bug where sometimes as I enter the passphrase to push via SSH it says "command already running". I have to ctrl+c out, and then try again, and then it works. Combined with the long passphrase, it's annoying to encounter this. Just generally want something that works FOR me, not just alongside me. Something that feels intuitive. And don't suggest tmux / zellij or similar, I don't want to jump around with standard keys. What I love in Warp is that I can actually scroll through vim files with the trackpad, jump to lines quickly without line numbers and other boomer shenanigans.
English
3
1
0
252
Ben Holmes
Ben Holmes@BHolmesDev·
@ejc3 Don't have to, but it's best to use smaller models when they have a high success rate at a task. Also good to use subagents when the context window is discard-able
English
0
0
0
5
Ben Holmes
Ben Holmes@BHolmesDev·
Computer use is so simple, but so powerful. - Set up a lightweight model that can run with low latency - Let it move, screenshot, evaluate, and repeat - Report back to the main agent when done with a final screenshot or video If you're annoyed with agents getting frontend wrong, this is the fix!
Ben Holmes tweet media
Warp@warpdotdev

Computer use is a huge deal. It lets agents close the loop by clicking around apps they build, verify changes e2e, and screenshot changes for review. Here's a technical deep dive from Daniel Peng (Warp eng) of how we built model-agnostic computer use for cloud agents 🧵

English
5
2
35
5.5K
Ben Holmes
Ben Holmes@BHolmesDev·
I’ve thought about making prompt guides or skill guides… but I honestly don’t find them useful. Just be a better communicator: - For agent orchestration, read product management tips - For in-the-loop development, learn from pair programmers or heck, psychologists
English
2
0
14
1.2K
am.will
am.will@LLMJunky·
I got so tired of everyone raving about how great cmux is. Panes this. Browser that. EXHAUSTING. And that's because I'm on Linux, where we get none of the coolest toys. So...I built it myself. And my God. You were right. It's amazing. Introducing Limux, a a GPU-accelerated terminal workspace manager for Linux, powered by Ghostty's rendering engine, with split panes, tabbed workspaces, and a built-in browser. Think cmux, but native Linux. If you're interested in something like this, be sure to leave a comment and I'll release it. Special thanks to @manaflowai and @mitchellh for making this possible.
English
70
20
432
42.7K
Ben Holmes
Ben Holmes@BHolmesDev·
@m13v_ @warpdotdev Good call on a11y trees. We wanted to start with a general-purpose solution before we dive into web app or native app use. Def optimizations to be gained. And yes, we show cross-platform diffs in our post! See "Keyboard input and the layout problem" warp.dev/blog/computer-…
English
0
0
0
17
Matt
Matt@m13v_·
@warpdotdev cool approach. on desktop we found accessibility APIs way more reliable than screenshots for targeting - you get the actual element tree so the model knows what it's clicking instead of guessing from pixels. curious if the action protocol handles cross-platform input differences
English
1
0
1
117
Ben Holmes retweetledi
Warp
Warp@warpdotdev·
Computer use is a huge deal. It lets agents close the loop by clicking around apps they build, verify changes e2e, and screenshot changes for review. Here's a technical deep dive from Daniel Peng (Warp eng) of how we built model-agnostic computer use for cloud agents 🧵
English
6
11
66
10.7K
Kuba
Kuba@kubabuilds·
@tobixion Thanks! I am using Warp as my terminal and played a little bit with opacity and backdrop blur 😁
English
1
0
2
12
Ben Holmes
Ben Holmes@BHolmesDev·
Nice, sounds like agent memory! Not something we have built-in yet, but you can hack your own by asking it to write down takeaways as markdown files somewhere on your computer. We also have "global rules" you can use to remind the agent to read and write those files docs.warp.dev/agent-platform…
English
1
0
1
27
David Silva Smith
David Silva Smith@DavidSilvaSmith·
@petradonka @warpdotdev had it help me with a couple things today! 🔥 Hoping there is some way to teach it things so I can be vague in the future and it will remember/lookup what I want to do, vs. doing research steps to figure it out.
English
1
0
0
7
Ben Holmes
Ben Holmes@BHolmesDev·
@KevinTrethewey @warpdotdev Understood. Was the issue here that Warp just helped with the commits without writing the code? Could also listen to rules better
English
1
0
0
257
K
K@KevinTrethewey·
Please stop this dark ux pattern @warpdotdev. You produce a great agentic tool (one of my favorites), but an agentic tool that doesn't listen is a liability, not an asset. Why would I pay for a liability?
K tweet media
English
1
0
0
314
Ben Holmes
Ben Holmes@BHolmesDev·
@aidenybai @warpdotdev Additional note: we decided to separate tab management from conversation history in ours, but it looks like yours is a simpler vertical tab model. Have a preference between the two?
English
0
0
0
63
Ben Holmes
Ben Holmes@BHolmesDev·
@aidenybai Love this pattern. cmux def inspired us to build a vertical conversation switcher into @warpdotdev to make this native, plus fuzzy find and a `/conversations` shortcut for inline switching. Gonna add Claude Code, Codex, et al support too x.com/zachlloydtweet…
Zach Lloyd@zachlloydtweets

Agreed. It's why we've been building a command center in the Warp terminal so you don't need to learn tmux to manage agents. - All conversations are searchable from history - Close tabs and revisit later with session restore - Track both local and cloud agent runs

English
1
0
1
563
Ognjen Gatalo
Ognjen Gatalo@ognjengatalo·
@unfiltered_ajit I’m using VS Code + Warp as my terminal. You can spin multiple claude code instances with multiple thinking levels without breaking your flow.
English
1
0
2
21
Ajit Gupta
Ajit Gupta@unfiltered_ajit·
Byee byee cursor!! Moving to VScode + claude code combo now Which one you are using??
English
64
1
79
4.4K
Ben Holmes retweetledi
Zach Lloyd
Zach Lloyd@zachlloydtweets·
We're exploring a /orchestrate primitive to let an agent spawn and talk to other agents. What about this UI to monitor a team? Review the delegation plan, then monitor your Agent Team with live status updates. Clicking will take you to a split pane for the subagent session
Zach Lloyd tweet media
Zach Lloyd@zachlloydtweets

We're exploring a new Skill to orchestrate agents automatically: /orchestrate The main agent writes a plan for how it'll delegate. When you approve, it spawns and manages subagents. You get full visibility of each subagent with split panes. No tmux to learn. Would you use it?

English
3
2
27
4.8K
Nick Gray
Nick Gray@nickgraynews·
@warpdotdev I used to get nice and friendly names for these tabs while using Warp with Claude Code, but no longer - is this a bug on my side?
Nick Gray tweet media
English
2
0
0
723
Greg (Autism Mode)
Greg (Autism Mode)@greg5figs5inch·
@levelsio Always wanted to do this, but never found a good way with Warp Terminal Might have to try to replicate this
English
1
0
0
112
@levelsio
@levelsio@levelsio·
Are you guys aware I am coding mostly on my phone now all day via Termius to Claude Code on my server while I go with gf to the dentist, clothing store, cafe, etc. 😛✌️
@levelsio tweet media
rootkid ✌️@rootkid

@levelsio "You" ➡️ IP your Internet provider assigns you; not your servers IPs. If you had a static IP I'd like to know why you prefer Tailscale over just adding e.g. your company IP to the firewalls SSH whitelist.

English
320
88
2.1K
677.8K