Ryan Ford

@8bitbandit @zachkrall @AnthropicAI Yes! I’ve essentially stopped using Dispatch unless it’s a quick/easy continuation of an existing task - Too much is lost in context and I end up spending more time going back to fix issues that were caused

@zachkrall Seriously @AnthropicAI this needs to be fixed as well as the dispatch disconnect to Claude code desktop. Why can’t I just use the mobile app and continue my Claude code thread I have on my local machine via Claude desktop

ok i still dont understand why chat, cowork, code are even separate tabs to begin with

@zuess05 College will become obsolete (at least compared to what it was just 5 years ago) stunningly quickly. That’s a good thing.

Genuine question. If Claude can now build production-ready apps in a few hours... Why are 18-year-olds still taking out $100k in student loans for a Computer Science degree? What exactly are they learning?

@victor_bigfield They can solve problems based on input data. Currently, the majority of data input is coming from the tech industry. Once you start challenging these models on fields it’s not as familiar with, the output quality decreases quickly. There’s still opportunities, but not forever

If anthropic and claude can solve all the problems that startups around the world are solving, what's the point of building saas?

@SidJain_80 Data quality / consistency

What’s the real bottleneck in AI systems? 🧠 Model quality 📊 Data quality ⚡ Latency 💸 Cost

@BillyM2k @iamSpacemann @jyoti_mann1 just posted an exculsive about Meta engineers competing on a company leaderboard to gauge who was using the most tokens - Makes you seriously question how smart some of these industry leaders are

@iamSpacemann many many companies

i think the whole “if you use more tokens you’re a better engineer” trend is about as negative iq as judging engineers on lines of code

@craigzLiszt lol, except the relationship is unhealthy, more like an addiction.

you’re not a serious developer unless you’re romantically involved with your coding agent

They have no excuse for not being a super developer - There biggest challenge is they are software experts, not industry specific workflow experts. They can build anything, but they don’t know what to build. They need to be partnering with industry experts (boots on the ground, not execs) to build products that automate real world workflows. The opportunities are endless.

Senior developers with AI are super developers. Agree ?

@KaiXCreator I disagree. There are still moats around complex workflows. This statement may be true in the future (possibly the near future), but not with currently (publicly) available LLM’s. Better question is - How do we differentiate in the future?

Founders, why are you still building? There’s no moat anymore. How are you gonna hit $1K MRR today when AI can clone your SaaS in seconds?

I’m new to developing with AI (dev in general, to be more accurate). I’m trying to do what I can to avoid “AI slop” code with structured workflows and test - What are your definitions of AI Slop? Best practices for limiting it?

Tips for Claude Token efficiency when building? I’m burning through Max usage in 2-4 hours. I already have .md standard workflow prompts and .md build plans (viewed directly by Cowork). I don’t use Sonnet or Haiku much outside of sidebar conversations/questions - I don’t feel comfortable using them for detailed build tasks, but I also can’t say I have a great feel for their performance vs Opus. I keep task session as short as possible. In fairness, I’m building two projects and often running 2-3 tasks at a time

@alexwtlf I’m hitting 20x limits in about 3 hours of continuous use. To be fair, I have 2 projects I’m working on and running 2-3 tasks at a time. I’m breaking everything down and typically using standard workflow prompts accessed directly by Cowork - Any other tips to save on usage?

Hitting Claude Code limits means you’re either building something insane or just broke

@iraftopo She might not have been wrong

@EngineLayer_ai Hehe... Yeah probably there's no point explaining after some age 😂 My grandmother when she was alive and tried to talk to her about tech stuff, she always told me... ..those things are the devil's stuff 👺

Good evening 👴 I’m your 80-year-old grandfather… Explain your product like I have no idea what the internet is 👇😁

@0xlelouch_ However, if you are self aware of your weaknesses, you can use it to help bridge the gap in those areas

Had a discussion with a colleague about AI. AI isn’t giving you skills. It’s amplifying whatever you already have. 1. If you write messy code, you ship bugs faster. 2. If you understand the domain, you ship features faster. 3. If you know debugging, you find root cause in minutes, not days. In this age, “learn AI” is not the plan. Go deep on fundamentals, then use AI to scale your output. But go deep, that agency and will to understand the depth of things is not going anywhere!

Good morning! I’m new here, and currently enduring an extended token time-out. How does this work?

@rxhit05 Yes

Can you call yourself a founder if your entire product was built by Claude?

@yegor256 It’s often simply a matter of asking the right questions

AI does not merely amplify expertise; it also compresses the gap by accelerating learning. The advantage is no longer defined by what you know alone, but by how quickly you can iterate, question, and adapt, making the speed of thinking as critical as depth of knowledge.

@Zinny_Edmund Virtually all startups need AI, but it can’t just be AI - The playing field is getting more level every day (everyone has AI, or at least access to it). Startups need a deep understanding of the problem they are solving

Do you think AI-first startups truly have a long-term advantage, or will the ease of access to the same tools level the playing field for everyone eventually? Why?

@AnythingBuilt How are you handling storage for the large pdf files? What’s the output? A narrative of changes?

@TTrimoreau Real world experience

If everyone has access to the same knowledge, what creates the difference?