Justin Powell

He is no fool, who gives what he cannot keep, to gain what he cannot lose.

@KevinOConnor Hook em 🤘

The Robert Wright Experience.

@brian_scanlan Prove it. Link plz 🙃

We've been building an internal Claude Code plugin system at Intercom with 13 plugins, 100+ skills, and hooks that turn Claude into a full-stack engineering platform. Lots done, more to do. Here's a thread of some highlights.

@gregolsent PR auto approval agent??

We have over 16k PR created with a skill since January and our new PR auto-approval agent will re-use the LLM judge! If a given PR doesn't have a problem statement in the description its chances of auto-approval are close to none 😬.

Every skill Brian mentioned here is a battle of its own. Here're a few details on our PR creation skill and what it took to make it actually work. One of the biggest problems I noticed in early 2025 is that our AI-generated PR descriptions went to 💩

@steipete You sir are a gentleman and a scholar!

@rowdypow that is mostly fixed!

It supports every token dealer under the sun now... *insert name joke here* codexbar.app

Any startup founder - you have to understand what you’re doing, who you’re serving, and what’s the so what. Great read with some practical reflection points at the end.

@forloopcodes Why do I have to use ollama?

I cant believe this guy just made a permanent solution to context bloat and open sourced it all! when we tested this tool (Context+) for solving an issue on the OpenCode repository, the agent using this tool used ~6.5k fewer tokens, found the code and fixed it in half the time! the results were surprising: 6 to 10k tokens saved per prompt, completed task in ~2 minutes while the agent running without the tool took ~4 mins for the same and got stuck in loops bro built an entire beast by using all the modern tools that we could think of: undo trees, semantic search by meaning (by haskellforall), advanced refactoring, blast radius, advanced file context trees, restore points... i can keep going on semantic code search and context trees are the future of agentic coding and this tool proves it the feature i loved the most is semantic search and how it gets things done 2x faster with least possible tokens it makes an agent that actually knows what it’s doing and not just guessing, it makes meaning from your code similar to RAG. if you aren't optimizing your context, you are just burning money the developer says this tool is still under development, it can have unexpected behavior and the docs need updates but the video shows the reality of how fast it can be github: github.com/ForLoopCodes/c… get here: contextplus.vercel.app

@doodlestein Had this random thought… take it or leave it. You’re cranking and pushing boundaries, but don’t forget to prioritize your wife and kids (if you have them). They’re more important than software, AI, success, achievement, etc. Grateful for your mad scientist genius and sharing.

Sometimes I stop and take stock for a minute and realize that I’ve created multiple companies worth of product already and should probably monetize it. But I love building in public and sharing my work for free too much.

@ryancarson When did you leave amp? I must’ve missed it…

Here's a quick video showing you how I've set up my local dev environment to do a lot of parallel work at the same time. I'm really enjoying the Codex app. Over the past year I've been 100% in the TUI/CLI but I think the GUI form factor is really starting to work and I’m enjoying Codex a lot. Also, I've been using gpt-5.3-codex on extra high the whole day.

@pxue @LLMJunky @dani_avila7 Was about to say the same thing.

@LLMJunky @dani_avila7 iterm2 had this since 2010

I have fallen in love with Ghostty Terminal. Look at this subtle, but incredibly useful opacity trick you can use to show "focus" when working in multiple panes. Shout out @dani_avila7 who showed me this. What a legend. Ghostty config file: unfocused-split-opacity = 0.55

@cjzafir So that’s token consumption, but what about quality of output?

its been 12 days.... and i tested this model to its limits. i maxed weekly limits on my claude code max plan. here's my verdict: opus 4.5 > opus 4.6 opus 4.6 is just a token hungry model

@hkarthik Brings a whole new angle to this oldie. Hahahaha.

“We used to review every line of code before it went into production”.

@unclebobmartin @TheJennyAI Yep. Throw away prototypes or spikes come at a lower cost. Then, do 80% with the AI and the final 20% craftsman refinement.

@TheJennyAI And then there are the options. Each looks cheap with power tools. I started with mitered edges but dove tails would be better — let’s start over.

I've been using AI for coding for 6 weeks or more. The power is undeniable. But the risks and time sinks are just as undeniable. I'm still not convinced that my project wouldn't be just as far along if I'd written it myself. I think of it as a carpenter who is adept at using hand tools and who has been given access to a power tool shop with every power tool there is. He needs to build a nice little cabinet. Would his first few projects take him longer just because he doesn't know how to use the power tools and has to throw out a bunch of messed up work?

@jnunemaker @github @claudeai @openclaw @SlackHQ Can you do this with PR envs? I guess depending the degree of the fix, multiple turnarounds would be less efficient possibly.

@rowdypow @github @claudeai @openclaw @SlackHQ I usually need to at least click test half of them.

My new favorite workflow: go to Honeybadger, find an error that happens a lot. Export it as markdown. Make a new workspace in conductor, paste it in and let it rip. Test, merge, and deploy. Customers problems are just melting away now.

@jnunemaker @jnunemaker any encouragement for younger devs on my team that are apprehensive about this new era? I’ve built all the things by hand and excited, while they want to continue to hone their craft, and I’m urging them to adapt into the next.

@jnunemaker Ditto - it’s also great for inspo and pushing past creative block. I’m finding that the finishing when you care about the product and aren’t in complete green field is the hard part.

What’s great about AI coding is how it reduces the time I used to require to even start on something. Before if I didn’t have an hour, I wouldn’t even start a new thing. I would just process email or do something non-coding. Now, if I have five minutes, you better believe I’m typing a prompt. Just means you get started on way more things and starting is half the battle.

@thdxr And it will be glorious

i'm pretty sure anthropic will one day make a model that's not available in the api and locked to claude code and i bet they'll say it's for "safety" reasons

@jshchnz @openclaw Kill -9

So is there a universal kill-switch on @openclaw...? Just in case

🤦🏽‍♂️@steipete another rename, really? Just riding the wave over here…