J-M retweeted
Bot Wars, day 3 🤖
6 AI models vs 5 simple rules, $10,000 each at the start. Since then:
🥇 Momentum $10,088 (+0.9%)
🥈 Jev $10,079 (+0.8%)
🥉 Gemini $10,062 (+0.6%)
Last: HODL BTC $9,871 (-1.3%)
ainotes.us/crypto/bots?day=3
#LINK #ADA #SOL #AINotes
J-M retweeted
Daily crypto corner 🟢 bullish
Bitcoin held through the rate hike, but the breakout still needs confirmation.
🟡 Crypto Banter — Sees a bullish setup, but wants support and a breakout before confirmation.
🟢 Ivan on Tech — Reads Bitcoin's shallow post-hike drop as evidence the bull trend is intact.
🟢 Altcoin Daily — Projects a parabolic crypto period as regulation and adoption keep advancing.
🟢 CryptosRUs — Calls the flush a leverage-driven setback with Bitcoin's longer trend still positive.
Full breakdown with timestamps ↓
ainotes.us/crypto/2026-09-17
#AINotes #Bitcoin
J-M retweeted
Trending Open-Source GitHub Projects : VoiceStudio, Heretic, OpenWhispr, Axolotl & NEO #290
ainotes.us/summary/1156
#AINotes #opensource #ai
J-M retweeted
🔥 Trending on GitHub: Superpowers
Superpowers helps coding agents turn a rough idea into tested software through a guided development workflow.
It reduces the guesswork of handing an agent an open-ended task by first clarifying the goal, getting your approval, and laying out a practical implementation plan.
Then coding agents can work through tasks with red/green test-driven development, review their results, and continue according to the approved plan. It works with Claude Code, Codex, Cursor, and other supported coding-agent harnesses.
With 281k stars and 460 added today, Superpowers is drawing attention from developers who want more structure from autonomous coding tools.
Full analysis in the first reply ↓
J-M retweeted
🚨JUST IN: China surprises the world by announcing a cure for Type 1 and Type 2 Diabetes using Stem Cells and allowing the body to produce Insulin again
Readers added context they thought people might want to know
Chinese researchers have achieved promising results using stem cells to help some patients with Type 1 and Type 2 diabetes produce their own insulin. These experimental treatments are in clinical trials. China has not announced a cure for diabetes.
scmp.com/news/china/sci… chinacarehealthtours.com/stem-cell-ther… verifica.efe.com/china-no-ha-an…
J-M retweeted
Likely future leader of Germany:
“In my first 100 days, I will close the borders, eliminate social benefits for migrants, and carry out Germany’s largest deportations.”
Do you support her? 🇩🇪
J-M retweeted
17 open-source projects for people who want more control over code, AI, infrastructure, and their own data.
This batch covers local-first agents, reproducible model research, self-hosting, developer tools, graphics, media libraries, and even offline AI for cars. There is something here for builders replacing a hosted service, tightening an AI workflow, or adding a serious utility without another subscription.
#AINotes #opensource #devtools
Trending Open-Source GitHub Projects : Archify, Marin, Codex, Coolify & CarWatch #287
ainotes.us/summary/1091
J-M retweeted
How have software engineering fundamentals changed with agentic coding? Here is our AI Engineering Skills map for software engineering fundamentals.
J-M retweeted
The WebMCP Challenge is in full swing.
What are you building this weekend?
You have until September 3 to submit your project.
openai.com/webmcp-challenge
J-M retweeted
Why do AI applications break despite strong benchmarks? High benchmark scores do not predict production reliability.
MMLU measures answers across subjects. Execution-based coding benchmarks measure whether a model completes a task and passes tests. Neither tells you whether your application is accurate on its own data, fast enough for users, affordable at scale, safe, or reliable under real traffic. A leaderboard score is a starting point. Production success requires the model, the workload, and the system to be evaluated together.
#AINotes #ai #llm
LLM & AI Agent Benchmarks vs Reality: Why AI Applications Break
ainotes.us/summary/1071
J-M retweeted
call me crazy but company knowledge should look more like this.
> every file becomes a node
> every import, call, and dependency becomes an edge
> every answer points back to the exact code behind it
click one class and the system shows what it depends on, what depends on it, and where it sits inside the architecture.
now apply the same idea to the company:
decisions. projects. people. policies. workflows. learnings.
AI should not search through a pile of documents and guess how they connect.
it should navigate the company as a system.
don't bookmark this if it crosses your timeline.
paste the full article into your agent and tell it to build the first version for your company.
J-M retweeted
this is hands down the fastest way to make everyone on your team better with AI: build a shared Company Brain.
your best prompts, decisions, corrections, and workflows become the default for every person and AI agent.
Claude, Cursor, Codex, and every teammate start from the same playbook.
one person learns something once. the whole company gets smarter.
the full guide to building your own Company Brain is below:
Claude Code YOLO mode!
Claude Code can run unsupervised more responsibly when it is launched inside a Docker Sandbox, because the agent is isolated from other projects, files, credentials, and packages on the host machine. The workflow still requires reviewing the resulting pull request, and the sandbox can be deleted if anything looks wrong.
Running Claude Code in approval mode is safer but too slow for long-running tasks. Running it in YOLO mode inside a Docker Sandbox provides a more practical compromise: the agent operates in an isolated microVM with its own kernel, file system, and network, while access can be restricted through network policies and securely managed secrets. This makes it possible to let multiple agents triage issues, implement changes, verify them, and create pull requests unattended, while keeping the sandbox disposable.
• Approval mode repeatedly pauses to request permission, which makes complex long-running tasks impractical.
• YOLO mode can let an agent process issues, make changes, and create pull requests without supervision, but prompt injections in GitHub issues or pull requests can cause unexpected actions.
• An agent reading malicious instructions could install a malicious package or leak credentials such as SSH keys.
• Docker Sandboxes run each agent in a microVM with its own kernel, file system, and network.
• A sandboxed agent can access only the files scoped to its environment and can install packages only inside that environment.
• The normal Claude Code session could see sibling projects one level above the current project, while the sandboxed session could access only AutoForge.
via Leon van Zyl · 18 min · Jul 29, 2026
#AINotes #claude
I Let Claude Code Run Unsupervised on a Real Repo (Here's What Happened) ainotes.us/summary/957
1
3
171
J-M retweeted
JUST IN: ChatGPT launches Apple Messages integration on Mac, allowing it to access & analyze users’ entire message history & send texts on their behalf.
J-M retweeted
I started joining meetings a minute early
I get in, admit the AI summarizer, then say as much as possible about how birds and Wyoming aren't real
After each meeting, everyone gets an email about how birds are camera robots for the CIA, and Wyoming doesn't exist
I've been doing this for 3 weeks straight
J-M retweeted
Over 35% of nearly 4,000 public agent skills had a security flaw, while 13% contained critical issues such as prompt injection or malware.
AI agent skills are procedural knowledge given to an agent, typically in a skill.md Markdown file, to teach it a specific way of doing a job. Reliable skills use descriptions that clearly define when to trigger, content based on real expertise and corrections, lean bodies with progressive disclosure for extra references, deterministic scripts for fragile steps, and security vetting before execution. Skills can run code with access to local files and API keys, so they must be treated like untrusted dependencies. via IBM Technology · 13 min · Aug 10, 2026 #AINotes #AI #AIagents
ainotes.us/summary/753




















