Meng Li
@mengli512.bsky.social
92 followers 8 following 1.8K posts
AI Engineer, I share AI insights at https://aidisruption.ai/.
Posts Media Videos Starter Packs
mengli512.bsky.social
#Alibaba's Xinliu team has launched #iFlow CLI, a free #AI terminal agent that outperforms Claude Code in benchmarks. It executes tasks via natural language and supports multiple Chinese models. Could this become your new coding assistant?

aidisruption.ai/p/alibaba-re...
Alibaba Releases AI Agent iFlow CLI
iFlow CLI: Free AI agent for developers. Run tasks in natural language, no coding needed. Top performance, zero cost.
aidisruption.ai
mengli512.bsky.social
#NVIDIA's #DGXSpark personal #AI supercomputer is now available, bringing data-center-level computing power to your desktop. With 128GB unified memory and 1 petaflop AI performance, it can run 200B-parameter models locally.

aidisruption.ai/p/dgx-spark-...
DGX Spark: Personal AI Supercomputer Launched
NVIDIA DGX Spark: Personal AI supercomputer with 1 PFLOPS performance & 128GB memory. Run 200B parameter models locally. Now available at $3,999.
aidisruption.ai
mengli512.bsky.social
Grok Imagine v0.9: Free, fast AI image & video creation with surprisingly loose limits. Two core templates + 10 prompts inside. Which style will you try first?
#Grok #GrokImagine
open.substack.com/pub/aiquill/...
Grok Imagine v0.9 Prompt Tutorial
Master Grok Imagine v0.9 AI image & video creation. Get free templates for stunning static images and 4K looping clips with generous moderation limits.
open.substack.com
mengli512.bsky.social
Cachier: Easily add persistent caching to Python functions with a simple decorator to boost performance. How do you currently handle repeated computations?
#caching #Python
open.substack.com/pub/pythonli...
Python Cachier: Persistent Memory for Functions
Boost Python performance with Cachier: Easy function caching for faster code execution.
open.substack.com
mengli512.bsky.social
OpenAI partners with Broadcom to develop custom AI accelerators and networking systems, aiming to deploy next-gen AI clusters by 2026. Could this challenge Nvidia's dominance in AI infrastructure?
#OpenAI #Broadcom
open.substack.com/pub/frontier...
OpenAI Announces Partnership with Broadcom
OpenAI & Broadcom partner to co-develop next-gen AI accelerators & Ethernet networking systems, powering future AGI ambitions.
open.substack.com
mengli512.bsky.social
Claude Code launches plugin ecosystem enabling one-click installation of AI assistants and workflows. How can developers seize early opportunities in this blue ocean market?
#ClaudeCode
aidisruption.ai/p/claude-cod...
Claude Code 2.0.12 Launches Plugin Feature
Claude Code 2.0.12 launches plugins: Add slash commands, subagents & MCP tools. Share custom workflows and boost team productivity instantly.
aidisruption.ai
mengli512.bsky.social
AI-written content often feels hollow. Fix: clarify goals first, use AI for research/structure only. Process: define → prompt → build knowledge → collaborate (human plans, AI structures, human refines, AI audits).
#Gemini
open.substack.com/pub/aiquill/...
Gemini: Better for writing, the more you use it
Stop hollow AI content. Learn a proven workflow to use Gemini for research & structure, not ghostwriting. Create substantive, high-quality writing.
open.substack.com
mengli512.bsky.social
#IREN surged 1000% in 6 months as it pivoted from Bitcoin mining to AI infrastructure. With NVIDIA partnership and 2.9GW power capacity, can this computing utility become the next CoreWeave?

open.substack.com/pub/frontier...
IREN: 10x Growth in 6 Months, Just the Start?
IREN: From Bitcoin miner to AI computing powerhouse. Explore its 10x growth potential and $100B market cap vision in the AI infrastructure boom.
open.substack.com
mengli512.bsky.social
Canadian team's DART drone achieves 80% success rate landing on vehicles at 100km/h. Can this breakthrough solve the 47% of drone crashes that occur during landing?
#Drone
open.substack.com/pub/robopub/...
Canadian DART Drone Hits 80% Landing Success at 100km/h
DART drone lands on 100km/h moving vehicles with 80%+ success rate using friction damping and reverse thrust tech.
open.substack.com
mengli512.bsky.social
Musk's Grok AI generates videos instantly, but "Spicy Mode" just strips characters. Can this gimmick really make his 2026 movie dream come true?
#Grok #video
open.substack.com/pub/aiquill/...
Grok Video Test: Fast but Flawed
Grok's new AI video tool: lightning-fast generation with a controversial 'Spicy Mode' that pushes creative boundaries.
open.substack.com
mengli512.bsky.social
High-quality training data is the key bottleneck in #AI development. DeepFabric revolutionizes this with structured, topic-driven synthetic data generation, enabling broader coverage and higher consistency.

open.substack.com/pub/pythonli...
DeepFabric: Next-Gen AI Training Data Engine
DeepFabric: Generate high-quality, topic-driven synthetic data for AI training. Boost model performance with structured, diverse datasets.
open.substack.com
mengli512.bsky.social
Bank of America calls #Palantir the "hidden champion" of #AI, raising the target price to $215. How is this government-contract-focused company leveraging AI Forward Deployment Engineers to break through amid bubble debates?

open.substack.com/pub/frontier...
Palantir: The "Next Big Winner" - BofA
Palantir: BofA's "Next Big Winner" with AI deployment secret weapon and $215 target. AI implementation leader beyond the hype.
open.substack.com
mengli512.bsky.social
Fei-Fei Li launches household robot challenge to advance embodied AI, with $1,000 and an RTX 5080 for the winner. Which do you think will become widespread first: household robots or self-driving cars?
#AI #BEHAVIOR
open.substack.com/pub/robopub/...
Fei-Fei Li Launches Robot Housekeeping Challenge!
Fei-Fei Li's BEHAVIOR Challenge: A benchmark for embodied AI to master household tasks. Join the competition by Nov 15, 2025.
open.substack.com
mengli512.bsky.social
Build a coding #AI #agent in 300+ lines that autonomously programs, debugs, and refactors. Just say "create a flask app" – it handles planning, coding, and execution while explaining its process.
aidisruption.ai/p/write-a-si...
Write a Simple Version of Claude Code in 300+ Lines!
Build an AI coding agent in 300+ lines! Automates programming, debugging, and refactoring using function calling and agent loops. Learn core AI agent concepts.
aidisruption.ai
mengli512.bsky.social
#Zhipu's new #GLM-4.6 just outperformed #Claude Sonnet 4 in real coding tests and leads Chinese models across 8 key benchmarks. Fully open-sourced via MIT license.

Would you switch your coding workflow to test it?

aidisruption.ai/p/glm-46-bea...
GLM-4.6 Beats Claude 4 in Coding, MIT Open-Source
GLM-4.6: Beat Claude 4 in coding tests. 27% faster, 30% less tokens. MIT open-source.
aidisruption.ai