Rich Low retweeted
what i’ve come to understand and hold true in my head but am always revisiting…
1. we don’t have enough raw resources (memory, cpu, power, data centers) - this won’t go away for perhaps up to 6 to 10 years
2. There is minimal to no moat for the prop labs; except selling poison and cure. They will make money selling expensive expensive cyber tokens in response to the security problems created by AI
3. AI fundamentally has changed work and there’s no going back. The box is open and businesses need fewer humans to run a businesses.
4. AI is an economic munition. If a country depends upon AI (and they will) and their access is suspended then a country will be economically crippled by such sanctions.
5. Openweight models are typically three months behind prop labs. They are a critical counter to $ant.
6. Countries must develop their own sovereign open weight models which codify their culture/customs to protect/encode their nations value structures to protect against cultural homogenization.
7. Distillation is competition and fair game. How it is done is should be the focus of conversation (not “if” it’s allowed) - ie distillation currently happens through theft/credit card fraud - that isn’t good!
Though tough thoughts am pondering about:
8. Training on pre-existing IP is perhaps not theft of IP. The same way that a human reading a book, seeing an idea, learning concepts and then remixing that concept isn’t a copyright violation. Mathematics isn’t IP/copyright theft.
Rich Low retweeted
paul rowsthorn cut a three-week product launch process to one day.
the $500-a-month SaaS tool went with it.
this is why i made Operator AI. real work. real numbers.
watch: youtu.be/LMYTl0uISLc
listen: open.spotify.com/episode/7iY…
Rich Low retweeted
Don't become cynical at all costs. Ignore haters. Don't respond. Don't give them even a drop of attention. Forget them.
Focus on people who create and on those who deeply care about peace. Connect them, build bridges, promote them all.
And hope for the best.
Rich Low retweeted
I’ve found it helpful when I’m in a bad season like I’m in right now to just focus on having a good day.
“A good day in a bad season.” It’s a bite sized victory.
String a few of those in a row and a bad season feels a little less bad.
Rich Low retweeted
Another bolt.new tip:
As a bolt.new builder, how do you plan complex projects when you've never built software before?
Here's what I've been experimenting:👇
Context: I'm building my 4th project on bolt.new. As it grew more complex, I needed a way to handle different parts of the app talking to each other - without burning through tokens and avoiding error-loops.
The Challenge:
As non-developers using AI to build apps, our biggest struggle isn't just writing code - it's managing how different parts work together - especially if the projects becomes larger.
Think: Your project has 15+ database tables with complex relationships. Now you need to update a frontend component - but the LLM might have lost context of your database structure. One small change could affect multiple connected parts, and LLMs tend to forget these connections when handling large projects.
The Solution:
I found inspiration in how GANs (Generative Adversarial Networks) work in AI. Think of GANs as a collaboration between an artist and an art critic – the artist (generator) creates work while the critic (discriminator) provides feedback, leading to continuous improvement through their interaction.
This got me thinking: What if we applied this same principle to building software?
Instead of trying to fit everything into one LLM's context window, I built a mini system inspired by GANs, where specialized LLMs work together, each handling different aspects of development.
First, let's understand how web apps are structured.
Core Components of Web Applications:
- Frontend (.tsx files): What users see and interact with (UI/UX) - your pages and components
- Backend (.ts files): Server-side logic and data processing - your API routes and services
- Database (.sql files): How data is stored and structured - your database schema and queries
- Business Logic (.ts files): Rules and workflows that connect everything - your utility functions and services
Next, I set up Claude projects to mirror these components:
1. Architecture & Planning LLM
- Creates initial project structure and file organization
- Designs database schema and relationships
- Plans API endpoints and data flow
- Maps user requirements to technical components
- Creates optimized prompts for bolt.new implementation
2. Frontend & UI Review LLM
- Reviews React components and pages (.tsx files) whenever there are issues
- Validates state management and data flow
- Ensures consistent UI/UX patterns
- Checks component reusability
- Verifies proper API integration in components
3. Backend & Data Review LLM
- Validates API routes and endpoints
- Reviews database operations and queries
- Checks business logic implementation
- Ensures proper error handling
- Verifies data validation and security
The Implementation:
I use separate Claude projects for this GAN-inspired setup - keeping the heavy thinking outside bolt.new.
Here's my workflow:
Feed each Claude project the current state:
- File structure from bolt project
- Functional requirement documents
- Database schema
- Current implementation details
Then follow this prompting strategy:
- Start with Architecture & Planning LLM to write/update functional requirements
- Take those to Backend & Data Review LLM to get proposed structures and workflow changes based on current implementation
- Finally, use Frontend & UI Review LLM to convert all this into optimized prompts for bolt.new
This lets me use bolt.new for what it does best: building the actual app.
Why such an elaborate setup? Well, that's the reality of building complex software without a development background! We may have to endure some pain of not knowing enough dev stuff, but bolt.new overcompensates by doing all the actual hard work! This system helps me think through changes systematically - like having a mini dev team guiding my decisions.
Why this works:
- Each LLM maintains specialized context
- Problems get caught early through continuous review
- Development stays clear and structured
- Save significant tokens in bolt.new
Results so far:
- 70% reduction in token usage (from ~1M tokens for 3-4 prompts to handling 10-12 prompts!)
- Faster development cycles
- Zero "everything broke" moments in last 2 weeks
It's fascinating to see how this GAN-inspired architecture scales - almost like having a tiny AI development team, each member specialized in their area. Makes me wonder about the possibilities if this was built directly into development platforms! 🤔
@stackblitz keeps making bolt.new more powerful each day. This structured approach to handling complexity might just hint at what's possible in the future of AI-powered development!
Important note:
This setup is overkill for small projects. I use it only for complex apps with multiple moving parts.
If you're building complex apps without a development background, hope this helps! Drop a reply if you need clarification or have questions.
P.S. Building something interesting with bolt.new? Would love to learn about your approach and how you handle complexity! 🚀
Rich Low retweeted
For the next 3 weeks I am going into BEAST mode 💪
🟡 Building my product
🟡 Building my body
🟡 Building my brain
Finding the way to X100 whilst building in public 🚀
Want to join the experiment?
Daily chat. No cost. No BS.
Repost this + send DM for personal invitation.
Rich Low retweeted
I'm giving away 5 premium license keys for 5 random people who help retweet the tweet below 🥰
Thanks!
Big Update: Sync Settings, 30+ new AI characters, UX improvements. 🎉
Let's take a quick look at our recent updates in the last 2 weeks:
🔸Claude Instant v1.2 is available
🔸Sync settings: choose which data to sync across devices
🔸Direct Plugin import: simply enter the plugin URL and import it directly into the app.
🔸Turn on/off plugins in the middle of a chat
🔸AI Avatar for AI characters + more than 30 prebuilt characters in our library.
🔸Download audio file from ElevenLabs
🔸Clear Context Button: added a button to clear context easily.
🔸 UI/UX improvements: changes in chat buttons, sidebar, and codeblocks.
🟢 New updates for TypingMind Custom custom.typingmind.com
▫️ Public mode: open access to everyone
▫️ Add tags to your members to control their access to the specific prompts / AI characters
▫️ Add training examples in Training data
▫️ Limit characters used for members in Usage & Limit
▫️ Post-message reminder to avoid hallucination
▫️ New admin dashboard for quicker access to the settings.
▫️ Custom theme for your chat interface
▫️ Custom HTML/JS/CSS code to your chat interface for live chat widget, analytics, etc.
Thank you for being so supportive. Stay tuned for more amazing updates!
--------------
🧪 Changelog for more update details: docs.typingmind.com/changelo…
📝 Use cases and tips: blog.typingmind.com/
Rich Low retweeted
I've spent 100s hours of finding good PM content.
Here's my HUGE curated db of 144:
🔸 tools
🔸 books
🔸 videos
🔸 courses
🔸 websites
🔸 podcasts
🔸 templates
🔸 newsletters
I'm going to sell it for $59.
For the next 24h only: FREE!
Follow + RT + comment "🚀" and I'll DM.
Rich Low retweeted
For folks in or aspiring to #productmanagement roles who don't have an engineering background: which of these topics would you most like explained in plain language, without any code? If the topic you'd most like to learn about isn't an option, please put it in the comments!
31%APIs
12%How the internet works
23%Components of an app
35%Algorithms
26 votes • Final results Rich Low retweeted
Hey current and aspiring PMs without engineering backgrounds! 👋 What do you wish you'd known about how software works when you got started, and what are you still wondering about now? #productmanagement #ProductManager #techcareers #SoftwareEngineering
Rich Low retweeted
Replying to @yanabana
Revenue and active users, absolutely. Admittedly, I work there, but I'm a longtime fan of Google's HEART framework for choosing the right metrics interaction-design.org/liter…
Rich Low retweeted
DUDE WTF engraved.blog/building-a-vir…


























