Finally, my skills are available as a Claude Code plugin!
claude plugins install mattpocock-skills
This means:
- No more manually syncing updates
- Aliases: mattpocock:code-review means no clashes with Claude built-ins
Perfect for folks tired of tinkering. Install once, never worry again.
Two senior Anthropic engineers just made Karpathy's famous agent loop far more powerful with one idea.
Graph Engineering.
They released an 11-page paper showing why agent performance improves dramatically once you stop thinking in loops and start thinking in graphs.
Here's the playbook.
step 1.
start with Karpathy's loop.
generate.
critique.
revise.
one self-review loop already beats a single model pass.
step 2.
give the agent tools.
search.
code execution.
databases.
reasoning without tools quickly turns into guessing.
step 3.
run agents in parallel.
give each one its own Git worktree.
same repository.
different branches.
no conflicts.
step 4.
connect everything with a graph.
instead of passing around conversations, agents write structured nodes and edges.
every claim stays linked to its source.
step 5.
ground the evaluator.
it checks claims against graph edges instead of relying on intuition.
if the graph can't verify it, the agent doesn't trust it.
step 6.
make the graph persistent.
the graph survives every session.
your agents stop rebuilding context from scratch and continue from where they left off.
this is the real shift.
Karpathy showed how one agent can think better through a loop.
Anthropic showed how thousands of agents can coordinate through a graph.
same models.
completely different architecture.
read the 11-page paper.
paste it into Claude.
bookmark this, then read the full guide on building graph engineering systems from scratch below.
ANTHROPIC JUST DROPPED THE OFFICIAL GUIDE TO PROMPTING OPUS 5
This is the most important prompting guide for anyone using Claude.
Bookmark this before you forget.
Most people still prompt Opus 5 like older models. That's the mistake.
> give the complete task upfront — don't drip-feed instructions.
> explicitly ask for shorter responses if you want them.
> use low/medium effort by default. increase effort only when needed.
> let it finish the entire feature instead of asking for partial implementations.
> use subagents intentionally for large codebases, not every task.
Opus 5 is at its best when you stop micromanaging it and let it own the work.
Read it before writing your next prompt.
Claude → Opus 5 → Agentic Execution → Production-Ready Output → Money
NotebookLM is now Gemini Notebook.
But the real upgrade is hiding behind the new name.
Use it like this:
→ Upload customer feedback.
→ Ask it to count repeated topics.
→ Generate a ranked table.
→ Build a chart automatically.
→ Turn the findings into your next content plan.
It no longer just summarises your files.
It can analyse them.
MCP protokolü bugün en büyük güncellemesini aldı. Kimse ne olduğunu tam anlamadı ama bu tarih değiştiren tarafta.
Eski MCP: her session'ın state'ini serverda tutmak lazımdı. Yani serverless çalışmıyor, edge'e atamıyor, load balancer arkasına sticky session'la bağlıyordun. Enterprise için can sıkıcı, indie geliştirici için pahalı.
Yeni MCP: stateless. Artık server'ı Cloudflare Workers'a, Vercel Edge'e, tek satır YAML ile Kubernetes cluster'a atıyorsun. Load balancer round-robin, sonsuz yatay scale.
Üstüne üç yeni katman geldi:
> MCP Apps: agent'ın kendi UI'ını sandbox iframe içinde render ediyor
> Tasks: uzun süren async operasyonlar
> Enterprise Managed Auth: identity provider'dan merkezi erişim kontrolü
MCP artık sadece "AI'ın USB-C'si" değil. Server'ı stateless, UI'ı render eder, uzun task'ları arka planda çalıştırır, SSO ile enterprise kilidini açar.
Yani her AI agent artık aslında bir mini SaaS platformu.
Vibe coder'ın hayali gerçek oldu: bir prompt yaz, agent hem hesabı hem UI'ı hem auth'u tek başına kurar.
Kaydet, 2026'nın en önemli teknik hamlelerinden biriydi.
Anthropic launched Claude Opus 5.
I plugged it into Agent OS immediately.
The difference was not just “better answers.”
It could:
→ Plan multi-step tasks
→ Check its own work
→ Find root causes in code
→ Use computer tools more reliably
→ Build interactive visual outputs
That matters when an AI is working beside other agents all day.
A smarter model is useful.
A smarter model inside a repeatable system is far more powerful.
Save this video, you’ll see why workflows beat isolated prompts.
The Full Guide is inside the AI Profit Boardroom. DM for the link. 💬