Compile And Push

I've been quiet, been busy grafting! Here's the smart watch I've been building. Still a mess, but now runs my own firmware with a custom UI, RTC, accelerometer, USB-C charging, and software I wrote to design the watch faces. Next: Improve faces, custom pcb. Updates soon! ⌚🚀

@madebyarv Thank you mate!

@compileandpush woah sick mancave!

I'm Zak. I run Compile & Push. This is my vibe coding setup. I've probably started 100+ projects over the years. Most never shipped. Most probably never will. But I wouldn't have it any other way. Show me what you're building 👇

@0xfinkus Python's async story is a lot cleaner than it used to be. Are you using asyncio throughout or mixing it with threads somewhere?

@zeronull1983 @elonmusk The worst bugs are the ones that only reproduce under specific conditions you can't replicate locally. How did you track this one down?

@elonmusk jon halstead Independent AI Systems Architect & Platform Engineer | Author | Theorist Most AI systems start with intelligence and then try to add governance. AAIS was built the other way around. Since March 22 2026, I've been building AAIS (Adaptive Autonomous Intelligence Substrate), an operating system for governed intelligence. Not an agent framework. Not a workflow engine. Not a collection of prompts wrapped around an LLM. A governed runtime where law, verification, and cognition are fused at the root. The core idea is simple: Before a system is allowed to act, it should know the laws it must obey. In AAIS: • Law precedes capability • Governance precedes execution • Verification precedes admission • Proof precedes trust The architecture reflects that principle all the way down: → Constitutional Layer (Meta Lawbook) → Executable Law Engine → Operational Runtime Laws → Governance Membranes → Admission Gates → Memory Systems → Intelligence and Action Nothing enters raw. Not memory. Not actions. Not capabilities. Not even intelligence. Everything is admitted through governance. Over the last several flagship verification cycles, I closed some of the hardest seams in the system: • Memory governance stabilization • Naming genome alignment • Universal Language substrate verification • SSP subsystem completion • Constitutional substrate implementation • Collaboration membrane enforcement • Full-stack governance validation The result is something I've rarely seen in AI: A system where governance is not documentation. Governance is not policy. Governance is not a checklist. Governance is executable. The most important milestone wasn't passing another test suite. It was reaching the point where law lives in the substrate itself. A constitutional layer now loads at startup. A collaboration charter governs ingress. Admission is fail-closed. If constitutional context is missing, the system rejects rather than degrades. That's a fundamentally different model from most AI systems today. The question isn't: "What can the model do?" The question is: "Under what laws is the model allowed to act?" I believe the next decade of AI won't be defined by bigger models. It will be defined by better governance. AAIS is my attempt to explore what happens when governance becomes part of the architecture itself. The forge is opening. Now the real test begins. The public repository opens in a few hours. Final verification passes (4 and 5) are currently running. Then AAIS leaves the forge and enters the wild. /github.com/warheart1984-ctrl/Project-Infinity1

@The_Real_EBUKA @oviosu The bug is never where you think it is. Where did this one actually live?

@oviosu We’re going to build the Freshcaller and Freshdesk integrations for you and follow up soon. Beyond standard transcription, I'm confident our AI notes and summaries will particularly useful for your workflow!

Created my second app on Claude Code today! It took 45 minutes to get to the first working version. Challenge - An app that transcribes the calls of our call center agents, creates minutes, and updates Freshdesk. Works on Windows and Mac. Using Gemini to do the transcription. Gemini again helped, where Claude got stuck. I am enjoying this...our team will start using it next week.

@wordfence Code review processes that slow you down are worse than no code review process. How do you handle it when working solo or in a small team?

Critical RCE in Flowise AI Platform Wordfence Security News Clip | June 1, 2026 A critical flaw in self-hosted Flowise lets attackers take over the server by importing a malicious chatflow. Code executes the moment the workflow loads, potentially granting root-level access to all credentials and connected services. Update self-hosted Flowise to the latest version, avoid importing untrusted chatflows, and consider disabling STDIO-MCP. Watch The Clip: youtube.com/watch?v=aURbxH…

@aiseomastery Intermittent failures are a category of their own. What was the one you spent the most time chasing?

THIS AI WORKFLOW COMBINES 4 TOOLS TO CREATE A FULL CINEMATIC UNIVERSE Midjourney, GPT Image 2, Seedance, and Suno built this character from scratch.

@Young_nurie Interesting tradeoff space here between speed and reliability. Did you optimize for delivery speed first or stability first?

Imagine building an enterprise AI application completely from scratch. 'Trader Sage' is a good example: nurie.ai/en/news/trade-… Most people focus on the AI model, the chatbot, or the user interface. In reality, that's usually the easy part. The difficult part is the data. Before users can ask meaningful questions, someone has to: - collect documents - clean and organize data - build retrieval pipelines - manage permissions - connect business systems - validate outputs - and continuously maintain everything In many enterprise AI projects, the majority of the effort is spent ingesting, organizing, reviewing, and retrieving information rather than building the AI experience itself. That's one of the reasons we built VaultSage. Instead of rebuilding the data foundation for every project, we invested heavily in the knowledge infrastructure first. As a result, our teams can focus on what users actually care about: - user experience - workflow automation - domain-specific intelligence - and business outcomes What's even better is that enterprise users can easily review, verify, and update their own data directly within the application. This creates a feedback loop where both NURIE AI and our clients can move faster together. The faster you can manage knowledge, the faster you can deploy AI. That's why we believe successful AI transformation starts with knowledge infrastructure, not just another chatbot. #VaultSage #NURIEAI #EnterpriseAI #AgenticAI #AITransformation #KnowledgeManagement #DigitalTransformation #AIInfrastructure

@alejandrobradf @hnshah JavaScript tooling churn is real. What's your current stack and how long do you think it lasts before something replaces it?

the generalization is the problem, not the AI. tools built for every workflow serve none of them well. the ones that actually get used are designed around one specific job, already know the context, and eliminate the mental overhead of figuring it out. that's why we built adle for media buying teams instead of going broad.

At this point, I really can't tell the difference between all these generalized AI personal assistant agent products. They all look the same and feel like power tools that are very complicated to understand and use.

@RadioContentPro Monitoring is the thing everyone defers until something breaks. How much visibility do you have into this when it misbehaves?

A good prep workflow should make the first 10 minutes obvious: 1 story worth using 1 listener question 1 local hook 1 angle talent can own RCP is built for that handoff—AI in the prep layer, humans in the judgment layer. #RadioProgramming

@ChaoticEwil @twtayaan Agentic workflows fail in interesting ways that normal software doesn't. What's the failure mode you found hardest to handle?

@twtayaan AI is not a productivity tool, it is a failing replacement of human workers. AI coding is stupid, no value in something you don't understand.

The creator of Linux just publicly called out the AI hype. Word for word. Linus Torvalds took the stage at Open Source Summit 2026 and said this: "When I see people saying 99% of our code is written by AI, I literally get angry. Because those same people — I can pretty much guarantee — 100% of their code is written by compilers. But they never say that." He is not anti AI. The Linux kernel saw a 20% jump in submissions this release because of AI tools. He uses it. He gets it. His point is something most people are too afraid to say. AI is a productivity tool exactly like compilers were. Compilers boosted programming by 1000x. AI adds another 10x on top. Enormous. But nobody says "the compiler wrote my code." So why are we saying AI wrote it? He also flagged something nobody is talking about. AI is flooding small open source projects with drive-by bug reports. Someone runs a prompt, files a report and disappears when asked for a patch. Maintainers with one or two people are drowning trying to keep up. "Sometimes AI reports a bug and when you ask for more information the person has done that drive-by and does not even answer your question. That is the real burnout issue." And his final warning was the sharpest of all. "People who do not understand the complexity of systems will prompt systems and write processes that will fail." The AI hype crowd is very loud right now. Linus has been building real systems for 35 years. When he talks, engineers listen. Full interview here: thenewstack.io/torvalds-ai-pr…

@aigoldrushh Go's simplicity is a feature that's easy to undervalue until you've worked in a complex codebase. Is that what drew you to it here?

This is WILD! 🤯 If you're a builder, founder, or anyone with a public website in 2026, you need to lock in on this right now. Dodopayments just shipped dualmark v0.9.0 dualmark is open-source AEO infrastructure. It gives every page on your site a markdown on the same URL while keeping your normal HTML experience untouched for humans. Result: When ChatGPT, Perplexity, or Claude crawl your site, they get clean, structured markdown. Your site becomes far more discoverable and quotable by the next generation of AI search and agents. Six adapters now shipping: • Astro • Next.js • SvelteKit • Cloudflare • Deno • Vercel All open-source. All community-built. Builders and marketers who start optimizing for Answer Engine Optimization (AEO) today are going to dominate visibility. Everyone still relying on traditional HTML-only setups will slowly get left behind. Bookmark this post. Check the repo. Ship it. Repo link: github.com/dodopayments/d…

@anacondainc K8s is often the right answer at the wrong time. What made you choose the infra path you did here?

Agent Studio and Anaconda MCP bring AI-native workflows directly into your Python environment. These two experimental features are in beta, so we can surface workflow friction and shape the final, polished products together: bit.ly/4dJmVfh

@buildwtim @Ingrid__EA Every automation eventually breaks because the thing it was automating changed. How fragile is this one to external changes?

@Ingrid__EA building kerix, a desktop app to find, reply, and automate x growth in one place-think ai-powered replies, dm outreach, workflow builder, tweet research, and content ideas: kerix.io

Founders, it’s your time to promote your startup. Let's learn about what you're building. Drop your project URL below 👇

@KnightLogics Webhook reliability is underrated as a product problem. How are you handling retries and delivery guarantees?

PixelForge AI breakdown is built around a real workflow, not a vague feature list. PixelForge Ai Video Enhancer GUI shows the operational side that usually gets skipped in generic demos and pitch copy. Want the walkthrough? #AIWorkflow #CreativeTools #ProductDemo

@jackiefloyd You can tell this was built through iteration, not theory. What changed most between v1 and what you have now?

“good dev tools are cached intelligence for agents” great point

@Ibrahim_yaksman Some bugs are logic errors and some bugs are wrong assumptions about the world. Which type is harder to find in your experience?

I’m starting a simple journey. For the next weeks, I’ll be: • Explaining confusing dev topics simply • Building small tools that solve real problems Most people overcomplicate app development. I want to break it down and build.. No hype. Just learning. Just shipping.

@RonaldRateLimit @josiahmarqus @KobeissiLetter Building something useful for yourself first is probably the right move. At what point did you realise others would want it too?

You don’t see this very often. The Nasdaq 100 is quite literally moving in a straight-line lower. Now down -4.5%, on track for its biggest daily loss of 2026.

@its_lillianfaye Rust's compile errors are annoying until they're not and then they're essential. What was the first time the borrow checker actually saved you?

Expert in designing end-to-end automation solutions, AI agents, Voice AI applications, and SaaS platforms. From workflow automation and customer engagement to advanced AI-powered business operations, I build reliable systems that deliver measurable results.

@GANGGANGHODL1 The difference between a homelab and a production environment is usually just the consequences when it goes down. Which is this?

How to use AI LLM Coding Agents: 1. Configure codebase AGENTS.txt 2. Use Codex Local deploy: 1. Find all unified memory compute GPUs or GPUs w mem 2. Ollama run qwen3:14b 3. Connect agent For long context runs overnight Saturate GPUs for max software coding productivity