Java Backend Developer | Building enterprise apps with Java EE ☕ | Real Madrid till the end ⚽ | University of Ibadan CS ’26|
Ibadan, Nigeria
Joined October 2023
- Tweets1.2K
- Following352
- Followers182
- Likes5.4K
26% of AI agent skills had vulnerabilities. 😳
We spent years learning not to npm install random packages, and now we're giving agents skills straight from strangers.
Scanners like this should be step one before any skill touches your machine. 🔐
Before you install that AI skill someone shared on X, scan it.
In an analysis of 31,132 AI agent skills, 26.1% contained vulnerabilities and 5.2% looked malicious.
NVIDIA open-sourced SkillSpector, the scanner it uses before publishing skills to its own catalog. 20.8K stars.
→ checks 71 attack patterns across 17 categories: prompt injection, data theft, privilege escalation, poisoned MCP tools and more
→ scans a GitHub repo, a URL, a zip or a folder
→ gives each skill a 0 to 100 risk score with a clear recommendation
→ fast scan with no AI, plus an optional deeper AI review
→ works for Claude Code, Codex and Gemini CLI skills
Install:
uv tool install git+github.com/NVIDIA/skillspect…
Scan a skill:
skillspector scan ./the-skill/ --no-llm
Apache-2.0.
Cloudflare had WAF rules live for the PHP bugs before they were even public. 🛡️
Me: finding out about vulnerabilities from a CVE tweet, three days late, while patching my pentest lab. 😅
Coordinated disclosure working this smoothly is the quiet win of the week.
Cloudflare has deployed default web application firewall (WAF) protections for newly disclosed PHP vulnerabilities before public disclosure. The rules use request and response signals to stop exploit attempts and prevent exposed data from reaching attackers. cfl.re/488w8dc
12,000+ reverse-engineered projects in one searchable repo, plus an agent skill to dig through it. 🔍
My weekend plans just got cancelled. My data plan too. 📶
Reverse engineering is the best CS class nobody puts on the syllabus.
holy sh*t, someone mapped 12,000+ reverse-engineered projects you can use today
cars, smart homes, games, AI clients and abandoned services. searchable in one free repo.
the collection:
→ awesome-reverse-engineered
12,869 projects, with an agent skill that searches by product, vendor or protocol.
→ games (3,036 projects)
emulators, decompilations and rebuilt engines. shadPS4 and RPCS3 included.
→ devices & hardware (3,028 projects)
tools for readers, wearables and connected hardware.
→ unofficial API clients (904 projects)
existing clients to check before writing your own integration.
→ AI tools & LLM clients (737 projects)
reconstructed clients, system prompts and tools for closed AI products.
→ cars (520 projects)
vehicle integrations, charging tools and openpilot.
→ firmware (472 projects)
OpenWrt, Tasmota and alternative firmware for supported devices.
→ file formats (436 projects)
readers, parsers and converters for files your app needs to understand.
→ network protocols (422 projects)
implementations you can inspect instead of guessing how a connection works.
→ service revivals (173 projects)
replacement services for products their vendors abandoned.
search your target → check the license and maintenance → test the existing implementation → build what's missing
bookmark the index for your next integration. send it to the friend about to spend a weekend rebuilding one.
Me: I'll just fix one small bug.
Also me, 4 hours later: rewriting the entire service, 3 new bugs, zero regrets.
Gemini: stitches three open datasets into a 3D human body you can zoom inside, beating heart and all. 🫀
Me: still debugging why my JSF page renders twice. ☕
Med students are about to learn anatomy way better than our faded textbook diagrams.
Unpopular opinion: Java isn't old, it's stable. Your favourite framework of the month won't survive the next 3 years. Spring Boot will still be paying salaries. 🧵 Agree or disagree?
An open coding model doing 125+ tokens/sec on an M4 Pro. 🔥
My M1, already running Eclipse and TomEE: "please, I have a family."
Still pulling it tonight if my data allows. Local agentic coding is getting very real.
Meet Flint from @Recursive_SI! A ~40B parameter MoE open model for fast, accurate agentic coding use case made to run on-device.
Flint can generate >125 tps on a M4 Pro Macbook Pro and >150 tps on a 64GB NVIDIA RTX Spark on short contexts.
Try it on Ollama today:
ollama run flint
Claude Code:
ollama launch claude --model flint
Shipping AI-generated code backed by proofs instead of line-by-line review. 🧾
As a guy who vibe codes and then squints at every diff like it owes me money, I'm very interested.
Formal methods going mainstream would be a wild plot twist for my CS degree.
Chopin + Proof-Driven Development Kit from @GitHubNext and @MSFTResearch enable teams to ship AI-generated code to production without reviewing it line by line, with stronger assurance than code review and testing alone.
Now in research preview ⬇️
Tech hack nobody told you: press Ctrl + Shift + T to reopen the browser tab you closed by mistake. It works even after 10 tabs. You're welcome.
AI models helping find critical flaws in PHP, the thing quietly running most of the web. This is the AI use case I'm most excited about. 🔐
Now the hard part: getting every forgotten WordPress site on shared hosting to actually update.
OpenAI models helped researchers—operating as part of our Patch the Planet initiative—discover critical security flaws in PHP, an open-source programming language used to run an estimated 70% of websites.
Over the past two months, we led a coalition of maintainers, researchers, and security providers to prepare fixes and protections ahead of disclosure. This duration was necessary because the fix was quite intricate and required a major change to the PHP engine. Every site using PHP can now safely upgrade their version to stay secure, in addition to applying mitigations from various security vendors.
Due to the severity of these flaws and PHP’s widespread use, we consider this bug to be one of the most impactful vulnerabilities in recent memory. It also marks a new phase, where AI can find security bugs with internet-scale impact.
This is an example of the Defender’s Window in action: to our knowledge, frontier AI was able to find this vulnerability ahead of attackers, giving defenders time to prepare.
openai.com/index/coordinated…
Readers added context they thought people might want to know
OpenAI may claim a role in the coordinated response and patching process. But the initial discovery and exploit development, however, occurred several months earlier by an independent researcher.
x.com/lcfr_hax/statu…
Open weights tell you what a model knows. Open-sourcing the inference stack behind your own API teaches everyone how to serve it cheaply. That's the real gift here.
Weekend reading sorted, if NEPA allows. ⚡
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
🔗 github.com/deepseek-ai/DeepI…
1/4
Junior dev: "It works on my machine."
Senior dev: "Then we ship your machine."
DevOps: *invents Docker*
Agents that resume after a crash and keep running for months.
Meanwhile my Spring Boot app can't survive me closing the laptop lid. ☕
Durable agents might matter more than the next bigger model.
Introducing 𝚍𝚞𝚛𝚊𝚋𝚕𝚎-𝚑𝚊𝚛𝚗𝚎𝚜𝚜. Run Pi, OpenCode, Codex, Claude Code, eve, fx, or your own harness durably on Vercel.
With 𝚍𝚞𝚛𝚊𝚋𝚕𝚎-𝚑𝚊𝚛𝚗𝚎𝚜𝚜, agents automatically resume work after failures and run for hours, days, or months. vercel.com/blog/making-every…
AI agents can now reverse engineer apps without the source code.
Meanwhile I can't reverse engineer the legacy Java service my team wrote in 2014. And I HAVE the source code. ☕
🚨 Episode 111 of The Intelligence Journal is here.
Perplexity Computer Can Now Reverse Engineer Anything 🔍🤖
Perplexity just added REA, short for Reverse Engineer Anything, to its Computer agent. Point it at an app, and it studies how the app works without the source code, then builds a working version of its own. 🔍
Its demos include Figma, Photopea and Audacity-style apps, and it’s free to try until October 31. ⚙️
Join us as we break down:
- How Perplexity Computer uses REA to take apart apps it has never seen the code for.
- Why AI could make copying software faster and cheaper than ever.
- Could AI replicate Eon’s intelligence? 🧠
🗓️ Date: Oct 20, 2026
⏰ Time: 3:00 PM UTC
📍 Set your reminder: xitter.cf/i/spaces/1jGXgBNlZovKZ…
Me: asking Claude to center a div.
Claude, on its lunch break: designing protein binders against scorpion venom. 🦂
We are not the same.
New on our Science Blog: designing protein binders against scorpion neurotoxins.
Scorpions sting about 1.2m people every year, and antivenoms are often costly or in short supply. Claude has made an early step toward finding new treatments.
anthropic.com/research/claud…
