Flutter Engineer | 3.5+ years | Built production fintech apps with 5M+ installs | Dart, BLoC, Firebase | Backend: FastAPI, Python, PostgreSQL | Open to work
If you are a @FlutterDev developer or vibe coding with Flutter, go and write this prompt to check how efficient your agentic setup is:
"Audit my project against https://t.co/wcV6eiK8hE and make sure the project is utilizing all Flutter & Dart agentic tools efficiently"
Learn how A2UI's client-side functions can help reduce latency and get you deterministic results in GenUI apps!
@redbrogdon breaks down client-side functions in a real world scenario. Discover how offloading math and business logic to local Dart code speeds up UI generation, integrates naturally with the GenUI catalog and system instructions, and works with the Dart developer experience (read: hot reload) that you're used to.
AI audio tools have gotten ridiculously good lately, and a lot of the best stuff is open source
Here are 10 repos worth checking out:
1. AudioCraft
https://t.co/Z72pFXAWXT
2. MusicGen
https://t.co/Z72pFXAWXT
3. Bark
https://t.co/UUWfCj6igl
4. Demucs
https://t.co/Dk9O8NLNIF
5. Whisper
https://t.co/ZACcAH7yEY
6. RVC
https://t.co/a9ZrZ7zf6J
7. so-vits-svc
https://t.co/IZDnEqUAyc
8. Coqui TTS
https://t.co/ZGyoXvFTtZ
9. Stable Audio Tools
https://t.co/jYtNLkQzRc
10. Essentia
https://t.co/vkltiLdC6n
You can basically build a full audio lab now without touching a closed platform
Ads retrieval has to narrow a huge candidate pool in milliseconds. Historically we couldn't afford ranking-quality models there — too expensive to pull all that ad data into the GPU.
Next-gen Andromeda fixes the real bottleneck: HBM bandwidth. With NVIDIA, we moved the compute inside the GPU — embedding gather, MoE interaction, scoring in one fused kernel, keeping tensors in registers/shared memory instead of HBM.
Result: gating went 5.3x from 2.4B to 10B+ ads/sec (~85% of peak HBM bandwidth, as fast as a dot product), index matching passed 100B ads/sec. End-to-end: 1B+ ads/sec per server, scoring every eligible ad — 100x over our 2024 system.
Ranking-level quality at ~300x fewer FLOPs:
https://t.co/R0k7fxIaoF
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5:
https://t.co/JIhXN7OSh1
- Excellent recs without any logins (state is kept in local browser session)
- Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads
- Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version
- HackerTube Longs ✊
A bit more on the process:
I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing.
Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
holy f*ck, someone put an entire reverse engineering degree on GitHub for free
7 repos, 560k+ stars, and in the right order they take you from a raw binary to agents that reverse engineer anything
each repo is the next layer of the one before
1. Ghidra -> how to read any binary
the NSA’s software reverse engineering framework, where everyone starts
↳ https://t.co/e0OTy789fk
2. radare2 -> the same skill in the terminal
a UNIX-like reverse engineering framework and command-line toolset
↳ https://t.co/fGlGSOjPXx
3. LLM4Decompile -> an LLM that decompiles
turns binary code into readable C, and the 9B v2 model reaches 64.9% re-executability on its own benchmark
↳ https://t.co/TtHCBVR9MT
4. r2ai -> an LLM inside radare2
LLM-based reversing for radare2, the moment AI starts reading the binary with you
↳ https://t.co/RdbzpTn7h0
5. ida-pro-mcp -> let the agent drive the tool
bridges IDA Pro to language models through MCP
↳ https://t.co/HMCygJJ7RE
6. reverse-engineering-assistant -> the same for Ghidra
an MCP server for reverse engineering tasks in Ghidra
↳ https://t.co/Ev3Oy82un3
7. REA -> agents that reverse engineer anything
35 target types, from native binaries and APKs to firmware, websites and Electron apps, with every finding backed by evidence
↳ https://t.co/SL7BkF3aRt
work through them in this order and you can hand an old app, a binary or a firmware image to an agent and get evidence back
the full list of what REA can reverse engineer below ↓
holy sh*t, someone just mapped 51,000+ paid AI apps you can reverse engineer today
chatbots, writing tools, coding assistants and image apps, and most builders have no idea it’s free
the drop:
→ https://t.co/ncVBRjHC5s (51,000+ AI tools · 9,000+ companies)
the index itself: every AI app you can pick from, the list of wrappers worth looking inside
→ https://t.co/baOMltBY3H (143.9k ★ · 34.8k forks)
system prompts, internal tools and model info from Cursor, Claude Code, Devin, Windsurf, Replit, Lovable and Perplexity
→ https://t.co/0YeHDC94pb (62.7k ★ · 10.3k forks)
publicly extracted prompts from Anthropic, OpenAI, Google, xAI, Microsoft, Perplexity, Meta, Mistral, DeepSeek and Cursor
→ https://t.co/A0l27CsZj5 (51.6k ★ · 10.5k forks)
extracted instructions of ChatGPT, Claude, Gemini, Grok and Perplexity
→ https://t.co/tnCGt4DQvc (15.0k ★ · 2.1k forks)
a collection of leaked system prompts from widely used AI services
→ https://t.co/ZPASKmw583 (116 prompts · 33 companies)
the same kind of collection, readable in your browser
→ https://t.co/UtPYOBM1aY
the formula:
pick an AI tool you pay for → read its public prompt → let REA show what it really calls → rebuild it or stop paying
Relying on SMS OTPs leaves onboarding vulnerable to fraud & latency.
Firebase Phone Number Verification enables direct carrier verification with a tap. Implement this Kotlin Android flow to protect infrastructure budgets while keeping onboarding frictionless.
LiteTune 0.7 can train a Decider for flutter_edge_ai, like Jev on the device
Two ways:
• EmbeddingGemma 2: many options, short text
• a small LM (Gemma 3/4, Qwen3): reads the whole dialogue; choice, yes/no or
score
https://t.co/AAXB4O9Nmo
https://t.co/nD5MuHaBWB
⚡Iterating at lightning speed! Gemini 4 Carbon is SOTA across key benchmarks and combines frontier capabilities with deep safety & security research.
Flutter is coming to @CopilotKit , and we’re building it.
We’re partnering with CopilotKit to create a first-class Flutter SDK built on the AG-UI protocol.
Teams will be able to bring agent-native experiences to iOS and Android from a single codebase, including in-app copilots that understand what’s on screen, generative UI that renders natively on mobile, and human-in-the-loop flows that work wherever your users are.
We’re taking it to Mexico City Tech Week, so come find us:
🔹 Microsoft AI Tour: joint sessions with CopilotKit at the booth
https://t.co/aqU1ioJA2e
🔹 Claude Community House: a hands-on workshop on building agentic Flutter apps with CopilotKit
https://t.co/z0fIDIJih2
If you’re building AI into a mobile app, or thinking about it, we’d love to talk.
Render Generative UI smoothly in Flutter by minimizing startup latency.
@redbrogdon demonstrates how to pre-generate and cache A2UI messages asynchronously using Dart Cloud Functions and Cloud Firestore.
Introducing Fish Audio Realtime STT (preview): speech-to-text that understands the conversation, not just the words.
It knows when to answer, when to keep talking, and when to stop.
Ranked #1 of 20 on Cekura’s STT benchmark (2.21% overall word error rate).
Built for voice agents. Try it now 👇
Kimi K3.1 Pro is here — a 3T-class open-weight model with a 1M-token context window and major upgrades to coding, 3D creation, and financial research and other agentic capabilities.
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
https://t.co/r0qFLdvd6c