gustaf

Realizing I run what is probably a most complex Claude Code setups in existence. 76 custom skills, 24 MCP servers, 10 hooks firing on every tool call, forges that launch 18 parallel AI sessions overnight. I have built an entire autonomous development infrastructure over the past three months. It is, by any measure, sophisticated. It is also, I discovered, wildly inefficient. Here is what a single day looked like when I finally audited it: 193 Claude sessions in one day. Not because I was typing into 193 terminals. Because every forge track, every subagent, every eval run, every RBI conductor cycle, every Ralphy task counts as a session. Each one loads the full context: a 367-line CLAUDE.md, 24 MCP server definitions, 76 skill descriptions, 20 memory files, 10 hooks. That context gets re-sent as cached reads on every single turn. The math is brutal. ~50-90K tokens on turn one of every session. Multiply by 193 sessions. Multiply by average turns per session. That is where the quota went. The deeper finding: I am in the top 0.1% of Claude Code complexity, and the tooling was not built for it. Anthropic recommends CLAUDE.md under 200 lines; mine was nearly double. They document 3-5 parallel agents; I was running 18. My Ralphy forge tracks were silently inheriting Opus ($75/MTok output) instead of Sonnet ($15/MTok) because a single missing --model flag. The fix was not to remove capabilities. Everything I built stays. The fix was lazy loading: move reference content to on-demand files, scope MCP servers per-project instead of globally, narrow hook matchers to only the tools they actually inspect. CLAUDE.md went from 367 to 184 lines. MCP servers went from 16 enabled to 6 global. Zero functionality lost. The lesson that keeps repeating: sophistication is not the same as density. A system that loads everything everywhere all the time is not more powerful than one that loads what it needs when it needs it. I knew this architecturally. I had not applied it to my own infrastructure. Building in public means showing the crashes, not just the launches.

@felixleezd @ADPList design

Introducing Design Stack. Today, we're releasing @ADPList's entire masterclass library ⭐️ 80+ videos on design, career, AI, product and more (100% free) from all our flagship conferences. Why are we doing this? We believe opening this full knowledge base will move our community in the fast changing AI era. These video courses have taught more than 10k+ people in design and tech. 🚀 Comment for "design" and I'll DM you the full library. LFG! 🤘

@itsolelehmann Cowork

i'm running a live claude cowork workshop for non-technical people on april 22 by the end of the 2 hours, you'll have a fully set up marketing system on your computer that: > produces a full week of content in one sitting, dialed into your voice so it sounds like you on your sharpest day > turns any marketing framework or post into a repeatable skill that claude runs on command for you > builds sales pages in minutes so you stop paying designers and copywriters thousands > schedules tasks to run while you sleep so you wake up to finished drafts, fresh ideas, and updated reports every morning > writes launch emails, newsletters, and sequences using the same frameworks behind my 6-figure product launches all click by click, on your machine, while i do it on mine here's everything that you get: • the full 2-hour live workshop where you build everything in real time • 16 personal skills that i built over 100s of hours for my own business • the complete recording so you can rewatch anytime • a self-paced course version of all the material • access to Claude Marketing OS telegram group this system runs 90% of the marketing behind my 7-figure brand doing 15M+ impressions/month and it's all yours come april 22nd comment "Cowork" and i'll DM you the link

@TheMattBerman @Meta @HubSpot @Calendly MONEY

this cowork agent knows which ads will print money 😱 here's the system that knows what meta doesn't: step 1: pull the raw data → pulls @Meta ads, GA4 + your source of truth (ie @hubspot, @Calendly or a csv) → assumes broken UTMs, disconnected tools, no data eng step 2: map ads by angle → agent re-groups every ad by MESSAGE family (teardown, save-time, founder-led, etc.) → your best message is usually scattered across 3 campaigns + 2 audiences. meta can't see that. → now you can answer the only thing that matters: which angle is producing $ step 3: score buyers not clicks → not CTR. not CPL. not ROAS. → bookings. signups. revenue. whatever moves YOUR business. → 4 questions every day: what's real, what's fake, what's leaking, what's underfed step 4: the 4 outcome truths → REAL WINNER: strong clicks + strong buyers = scale → FAKE WINNER: strong clicks + garbage buyers = cut. no hesitation. → LEAK: good ad + broken page = rebuild page → UNDERFED: quiet crusher at 9% of budget = feed it step 5: find what's breaking → tells you exactly where it breaks: ad, page, or follow-through → last week it caught a founder-psychology ad landing on a generic SaaS page converting 34% below account average → not "something's off." but specific. actionable. step 6: 1 minute brief → 60 seconds brief. lands in cowork/telegram → every rec ships with a confidence tag. low data = no rec step 7: memory that compounds → week 1 finds the obvious fake winners → week 4 it pre-flags curiosity hooks before they become the next fake winner → every cycle is smarter than the last input: meta account + GA4 + one outcome source output: real winners. fake winners. leaks. what to do next. the old way: → triple whale: $300/mo → northbeam: $500/mo → hyros: $500/mo → data analyst: $8K/mo = $9,300/mo of "it depends" this: $0 I packaged the entire system as the Outcome Kit. 3 claude cowork agents + 12 skills: - data reader (meta + GA4 + outcomes. source truth only) - diagnostician (angle mapping + 4 outcome truths scoring) - brief writer (concrete moves + pattern memory) also works with @openclaw and hermes (@nousresearch). giving it away free. comment MONEY + like + follow (must follow so i can DM)

@adriansolarzz following! pls send

@_gmpalm follow, then I will send it over

just put together a 57-page page breakdown on how i test 300k+ AI UGC variations a month using Claude Code same multi-account distribution system that generated $800k+ for clients and for 24h, it's yours for free like + comment "TEST" and i'll send it (must be following + RT for priority access)

thats whats up

@SolAndrew_ yo

I have 23.1K verified followers If I'm still not following you, comment and I'll do it.

how are people dealing with SYCOPHANCY? learning to build genuine trust with claude?

@jordan_ross_8F marketing

Anthropic ran their entire marketing operation with one person. $380 billion company. Paid search. Paid social. SEO. Email. App stores. One non-technical hire doing all of it — for 10 months. I pulled it apart. Compared it to every system we've built across the clients we've worked with. Then asked myself one question: If I had to reverse engineer this from scratch — what would it actually look like? Turns out the architecture isn't that complicated. I mapped the whole thing into a 47-page PDF you can upload directly to any LLM. It coaches you through building your own version step by step. Comment "marketing" and I'll send it over.

Today we're excited to announce NO_FLICKER mode for Claude Code in the terminal It uses an experimental new renderer that we're excited about. The renderer is early and has tradeoffs, but already we've found that most internal users prefer it over the old renderer. It also supports mouse events (yes, in a terminal). Try it: CLAUDE_CODE_NO_FLICKER=1 claude

working on developing deeper trust with claude. open to hear others' wisdom

@NathanielC85523 @RampLabs honestly pretty consistent...

@_gmpalm @RampLabs have you seen the token gap widen on multi-step workflows specifically, or is it consistent across session types?

Today, we're releasing Ramp CLI to let agents manage your company's finances. 50+ tools across cards, bills, expenses, travel, and approvals. Fewer tokens than MCP, and comes with pre-built skills like receipt compliance and agentic purchasing.

@_chenglou What was used to make this?

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

@AhmedShahnab What was used to make this?

Nothing Fancy... just experimenting and exploring with the available ones before exploring the new trending ones.. The core trick here: ctx.measureText() pre-computes every word into a data object with a baseX, current x, and targetX. Each frame, a 120-segment dragon emits collision nodes with anatomically-computed radii. For every line of text, it projects each node onto the line's Y axis, compute the X exclusion span via √(r² - dy²), merge overlapping intervals, then run a relaxation solver (forward + backward pass) to find stable non-overlapping target positions. Words lerp toward those targets each frame - giving the parting/closing flow you see. Sometimes i feel platform APIs that's already shipped in every browser are still criminally underexplored. #CreativeCoding #Canvas #WebDev #TypeScript #InteractiveDesign #GenerativeArt #Frontend #TechArt

@claudeai @grok I'm with you @grok but I'm trying to build on my machine to more sophisticated and powerful infrastructure, how do I reconcile that?

I need @claudeai power users to reality check this what do you think @grok ?