Ankur Toshniwal

@garrytan 10 hours is too long no? I would've thought you'd have gstack automatically handling these things by now

GStack just shipped natural triggers so it'll help you do the things you want to do and you don't have to remember the skill names! Thanks to Mark Thurman on the YC Software team for this idea Suggested at 11:30am, shipped by 9:08pm same day

@SherryYanJiang Us. gopeek.ai

who are some peeps doing cool stuff around memory?? top of mind i could think of are zep, mem0, supermemory

@christiangori96 kick rocks according to Dario

If everything can be automated with AI, what will humans do?

@CharlieSandor @usehercules What is a CC analyzer?

I use Cursor / Codex for most of my longer running projects, but it is still annoying to tackle small side projects. On the flipside, @usehercules can basically 1 shot most apps in <15 minutes. I built a CC analyzer and saved myself ~$7k/year by cancelling unused subs (1/2)

@fashiongiik @Lin142535Lin dang, I'm new to X. didn't realize this was so prevalant. thanks for the heads up

@its_unks @Lin142535Lin Feel like an AI account, the emdash…

So I found this tool called gstack and it's kind of insane It basically turns Claude Code into a whole team of specialists. Instead of one AI helper, you get 13 different "modes" you can summon with slash commands. Need to think like a CEO about your product? `/plan-ceo-review` Want someone to tear apart your code for bugs? `/review` Need to run QA and actually fix the bugs? `/qa` Just want the AI to browse your app and take screenshots? `/browse` It's like having an entire dev team in your Claude Code session. Each command switches the AI into a different mindset — founder, engineering manager, designer, QA engineer... Setup is super simple: `git clone github.com/garrytan/gstac… ~/.claude/skills/gstack && ./setup` If you use Claude Code daily, this is legit a game changer. No more asking "can you review this like a senior engineer?" — just type `/review` and it happens. Been using it for a week now and my workflow is completely different. #AI #DevTools #ClaudeCode

@adisingh that's a translator lol!

Did an onboarding today with a (very nice) nontechnical guy who would ask me questions from and type my responses back into ChatGPT live. This is the future - humans talking to agents through other humans lol

Finding out that we built an entire eval product and the best part of it was literally just Agent Replay: #replay" target="_blank" rel="nofollow noopener">docs.simforge.goharvest.ai/python-sdk#rep… 1-line, replays production traces locally. Set it up on my search algo, looped CC on it and it tuned my fusion search parameters on: gopeek.ai

@big_duca unreal

“Dude did you vibe code this slop? This feature sucks!” Been getting this more recently. And no, I didn't “vibe” it. Did you ever consider, for one single second… That I might just be retarded? And I wrote this organic slop myself?

@Lin142535Lin I meant how do you keep your markdown files updates as your needs evolve? Do you use a separate skill for it? How do you get CC to learn your judgement/taste/goals?

@its_unks I spend a lot of time on Twitter, and I follow many people in the AI space, which has led to this.

@zeeg I meant the most recent one (Opus 4.5/6). Either way, building good skills and systems does go a long way. No idea how you're supposed to scale that to a a large org though.

@its_unks yeah theres def some kind of pattern to this - for me it was like 6-9 months ago, but you go through the wonders of how magical it is, then constant reality checks

i can write 50k lines of code a day and it will absolutely not generate any tangible lasting value nor will these 16k, from Garry or anyone else (sorry, but its the truth)

@justic_hot @calebsandovai @garrytan still no dice on this. We tried distilling, but we have a lot of business context we want to provide.

@its_unks @calebsandovai @garrytan huh really? mine follows CLAUDE.md pretty well. the trick was putting it at the project root not in some nested docs folder

The thing I believe that few people believe but I think everyone will believe Markdown *is* code

@garrytan Are these LOC going into gstack or something else? Is there a way to try any of the other products you're building?

I am coding a lot, GStack is helping me do it, but also I want you to know I was stranded in Austin the last 24 hours due to weather, and also last week my mom was in the hospital and not too lucid for most of it, so I was coding by her bedside too. She's ok now and I just visited her at home and set up her medication. I do have a full time busy job, and is it really possible for a CEO to be coding all the time? Frankly, I think it will have to be. The CEO has to set the future of the company. All companies will need to adapt to a faster world and do more. Boil the ocean. It's not about doing less and cheaper. It's about doing more and making 10x better products and services. Is 16k LOC/day sustainable for me? We're going to find out if I can manage to get to L8 software factory. I have not done it yet. But you can tell the models are about to get much much better. L8 is barely possible today, and I think I'm close. But everyone will be there soon. I want to be one of the people who helps all of you do it with me.

Word of the year to looking like a bot/sheep on any platform right now seems to be "quietly". Someone give me a filter.

@justic_hot @calebsandovai @garrytan herm... we have a monorepo. I will try this though thank you!

There's a decent chance that plugins are the earliest form of the iPhone app store. The marketplace itself is littered with bugs, but the utility of plugins is extremely powerful. Plugins >>>>>>> MCP or CLIs. code.claude.com/docs/en/hooks

@ud_ambwani Does CC now work better? Using MEMORY.md too?

spent way too long manually writing CLAUDE.md files. then realized i could just ask claude code "what context do you keep needing from me that i forget to give you?" and it basically wrote its own instruction manual. 10 min vs the hour i was spending

All I wanted to do was view my files on a separate worktree. Cursor and Conductor both made feel extremely stupid. Why does Conductor wrap my worktree names in a separate unrecognizable name? how does this help?

@y_nabih @garrytan Tend to agree-ish. LLM as a judge works well when it's trained (GEPA style) on ground truth labels. Almost no one goes far enough.

@garrytan Fully agentic" is just VC delusion for a state machine that hallucinates until it runs out of tokens. LLM-as-a-judge is circular logic. If you're still relying on "prompt engineering," it's a brittle demo, not an agent. Let's see if GStack brings math or just vibes.

Full evals system coming to GStack shortly. LLM evals are the only way you can make fully agentic systems that are repeatably better as you improve the workflow, context engineering and prompts.