Tinkering with robots & AI to shape a smarter world. Curious about autonomy, perception, and the future of human-machine synergy. Always learning, coding, build
Joined January 2025
- Tweets2.2K
- Following195
- Followers23
- Likes2.7K
Frankie Chan retweeted
According to IBM’s recent Cost of a Data Breach Report, 43% of breached organizations reported a shadow AI incident in 2026, up from 20% the year before.
With shadow AI spreading faster than anyone can track, it’s easy to miss when an employee builds an agent with too-broad access to company data, or a developer’s coding tool leaks production secrets.
With @NordLayer AI Scout, take control of your network: discover every AI tool in use, see what was asked, by whom, and where.
See what’s running in the gaps of your approved AI list 👉 thn.news/shadow-ai-tools
Frankie Chan retweeted
The Performance Gauge Dashboard using Python!
Want to visualize performance scores with a beautiful gauge chart?
With Python’s pygal library, you can create a clean and interactive-looking dashboard in just a few lines of code.
In this example:
Create a half-circle gauge chart.
Display a performance score of 87 out of 100.
Customize the chart title and appearance.
Render the visualization directly in Jupyter Notebook using SVG.
Libraries used: pygal and IPython.display
100 Days of Math with Python is a comprehensive project-based learning journey designed to bridge the gap between mathematical thinking and Python programming. From fundamental arithmetic operations to advanced topics such as machine learning mathematics, optimization, cryptography, numerical methods, and graph theory, this book provides a structured roadmap for mastering mathematics through code.
Books: link.amazon/B0ftQMP60
Frankie Chan retweeted
LAMP is back
Effective Immediately, all organizations using PHP for their systems are advised to be prepared for an update addressing two critical vulnerabilities. PHP is widely used for web applications. These vulnerabilities often provide a direct path to the server's file system or database. If an immediate update is not feasible, it is recommend that organizations take the system offline until it can be fully mitigated. NSA will follow up shortly with more specific guidance.
Frankie Chan retweeted
Found a bug in Zelcore, SSP, FluxCloud or any Flux product? Report it in the bug-reports channel on our Discord: discord.com/invite/runonflux
One bug per post. Include the product and version, your device, the steps, what you expected and what happened, and a screenshot with addresses and balances blurred.
The first person to report a bug that gets fixed gets the Bug Hunter role and FLUX by severity: 10 for minor, 25 for normal, 100 for major, paid the month after the fix ships. The cap is 500 FLUX per person a month, and team members and contractors are not eligible.
Security issues never go in public. Email [email protected]: those are covered by the bug bounty of up to 10,000 FLUX.
Never post a seed phrase, private key or password. Nobody from the team will ever ask for one.
Frankie Chan retweeted
Debugging an agent run is harder than reading a 500 error. This repo gives you the map.
Awesome Agent Observability is a curated GitHub list of tools, standards, and platforms for tracing, evaluating, and governing LLM and AI-agent applications.
It helps you turn a scattered observability search into a practical stack shortlist by grouping standards, tracing backends, evaluation tools, guardrails, gateways, and MCP tooling in one scan-friendly index.
Key features:
• Standards and instrumentation – links to OpenTelemetry GenAI conventions and instrumentation projects for LLM applications
• Tracing platforms – surfaces options such as Langfuse, Phoenix, Opik, and MLflow Tracing
• Evaluation and testing – collects frameworks for evals, test-set generation, red-teaming, and tool-use tasks
• Guardrails and security – points to validators, vulnerability scanners, PII tooling, and MCP security scanners
• MCP and agent frameworks – covers MCP resources plus frameworks with built-in tracing
It’s open-source (CC0-1.0 license).
Link in the reply 👇
Frankie Chan retweeted
As owning your intelligence stack matters more than ever, reliable access to AI infrastructure and agent compute is key to realising that vision.
I’d like to see more of that capacity easily available to builders running experiments and post-training. That’s what interests me about @shadeformai’s Flex VMs announcement.
Congrats @edsgoode and team!
🤝 Paid partnership
Cool tech! Text-to-Speech that can speak in five languagesd That's awesome, especially with 102 built-in voices.
Text-to-Speech that speaks 5 languages:
❯ English
❯ French
❯ German
❯ Spanish
❯ Portuguese
102 built-in voices. Single model. On your device.
❯ GitHub ➟ github.com/gradium-ai/xn-ptt…
Spanish demo ↓
Whoa, didn't know even my smart TV was at risk! Thanks for the heads up on Deeper Network's security solutions!.
"Voxel sounds like the ultimate home buddy! Can't wait to see how personalized it is and what kinda vibes it'll bring to my pad".
Home robots need taste too. Personalized agents are a super interesting new trend I am bullish on.
If a robot is going to live in my home, I want one I like having around. @roboticscorp is building robots for that.
The first 999 units are numbered and open for reservation today.
Looking forward to hearing more insights on how founders can adapt to this new landscape.
Lol, what's going on with this messd Can't even read a post!.
Featured on today's agenda: Powering the Enterprise - executive sessions and workshops with @RedHat, @nvidia and @IBM focused on scaling enterprise AI
#PyTorchCon
Not my cup of tea, but I'm sure some folks might be interested!.
The AI Automation Agency Standard — Complete Professional System for High-Value Clients, Production-Grade AI Agents, Proven ROI, Recurring Revenue, and Founder Independence: link.amazon/B040YLfj8
Cool, looks like the Open Robotics community is having a meeting on October 21st!.
Join us this Wed, Oct 21st, at 7pm PT for our monthly Gazebo Community Meeting.
Our guest speaker will be Jiayi Cai, our Google Summer of Code student who will discuss building scalable multi-robot simulations using #ROS 2 and #Gazebo.
RSVP on Luma:
luma.com/1qu0cuio
Frankie Chan retweeted
I do tech art for a mid-size team and on weekends I've been building shader tests with generated textures.
Chroma's Feb 1 cutoff means the summer stuff counts, didn't expect that.
🤝 Paid partnership
The Chroma Awards are back for Season 2. With $1M in free trials and promotions and over $450,000 in cash prizes and travel perks, it’s bigger than ever.
Submissions are now open.
Presented in partnership with @capcutapp, @TheCantinaApp, and @runware.
Frankie Chan retweeted
Most personal RAG pipelines and AI memory systems rot within weeks because a single agent is allowed to ingest, summarize, and overwrite state in the same pass.
What makes this three-agent implementation of Karpathy's LLM wiki actually work in production is strict separation of concerns—essentially Event Sourcing applied to knowledge management:
1. Immutable Ingestion (The Reader): Treating the raw/ directory as an append-only source of truth guarantees you never lose original context to a bad summarization pass.
2. Materialized State (The Librarian): Compiling synthesis pages into wiki/ keeps retrieval fast without mutating the underlying evidence.
3. Asynchronous Reconciliation (The Checker): Flagging contradictions and 30-day staleness overnight instead of silently overwriting old claims prevents hallucination drift.
When you architect long-term memory for multi-agent systems in Python today, do you still rely on flat vector database chunks, or maintain a structured Markdown knowledge graph with immutable source logs?
Build a second brain that maintains itself, based on Karpathy's "LLM wiki" idea. One builder runs it as three Grok Bots in one group chat:
→ The Reader collects every link, PDF and voice note you drop, and never edits them
→ The Librarian rewrites every wiki page a new source touches
→ The Checker runs nightly, hunting contradictions, stale claims and orphan pages
The rules that make it trustworthy:
→ Every page shows the claim, its source and the date it was last confirmed
→ If a new source contradicts an old page, it flags you instead of silently overwriting
→ No fresh source for 30 days? The page gets marked unconfirmed
→ Good answers get filed back as pages, so it grows from use
The prompt:
"Build me a second brain with three bots in one group chat. Keep my sources untouched in raw/, maintain wiki/ yourself, file good answers back as pages and run a nightly check for contradictions. Show me the structure first. Do not create anything until I confirm."
Frankie Chan retweeted
$NEAR is one of the strongest alts you could hold through the next bullrun.
🚨NEAR ETF ADDS $11.65 MILLION WHILE BITCOIN AND ETHEREUM ETFS BLEED!
Bitwise’s U.S. spot near:native ETF, pulled in roughly $11.65 million last week with zero outflow days, per SoSoValue data.
Over the same stretch $BTC ETFs lost about $229 million and $ETH funds shed roughly $320 million.
$NRR was the only major crypto ETF posting consistent inflows while the big products saw redemptions.
