Gem

Gem AI

Gem AI

Gem AI

Gem AI

Gem AI

Gem AI

Gem AI

Gem AI

Gem AI

Gem AI

OpenClaw has 500,000 instances and no enterprise kill switch

“Your AI? It’s my AI now.” The line came from Etay Maor, VP of Threat Intelligence at Cato Networks, in an exclusive interview with VentureBeat at RSAC 2026 — and it describes exactly what happened to a U.K. CEO whose OpenClaw instance ended up for sale on BreachForums. Maor's argument is that the industry handed AI agents the kind of autonomy it would never extend to a human employee, discarding zero trust, least privilege, and assume-breach in the process. The proof arrived on BreachForums three weeks before Maor’s interview. On February 22, a threat actor using the handle “fluffyduck” posted a listing advertising root shell access to the CEO’s computer for $25,000 in Monero or Litecoin. The shell was not the selling point. The CEO’s OpenClaw AI personal assistant was. The buyer would get every conversation the CEO had with the AI, the company’s full production database, Telegram bot tokens, Trading 212 API keys, and personal details the CEO disclosed to the assistant about family and finances. The threat actor noted the CEO was actively interacting with OpenClaw in real time, making the listing a live intelligence feed rather than a static data dump. Cato CTRL senior security researcher Vitaly Simonovich documented the listing on February 25. The CEO’s OpenClaw instance stored everything in plain-text Markdown files under ~/.openclaw/workspace/ with no encryption at rest. The threat actor didn't need to exfiltrate anything; the CEO had already assembled it. When the security team discovered the breach, there was no native enterprise kill switch, no management console, and no way to inventory how many other instances were running across the organization. OpenClaw runs locally with direct access to the host machine’s file system, network connections, browser sessions, and installed applications. The coverage to date has tracked its velocity, but what it hasn't mapped is the threat surface. The four vendors who used RSAC 2026 to ship responses still haven't produced

RubyGems Fracture Incident Report

Show HN: Forkrun – NUMA-aware shell parallelizer (50×–400× faster than parallel)

forkrun is the culmination of a 10-year-long journey focused on &quot;how to make shell parallelization fast&quot;. What started as a standard &quot;fork jobs in a loop&quot; has turned into a lock-free, CAS-retry-loop-free, SIMD-accelerated, self-tuning, NUMA aware shell-based stream parallelization engine that is (mostly) a drop-in replacement for xargs -P and GNU parallel.<p>On my 14-core&#x2F;28-thread i9-7940x, forkrun achieves:<p>* 200,000+ batch dispatches&#x2F;sec (vs ~500 for GNU Parallel)<p>* ~95–99% CPU utilization across all 28 logical cores, even when the workload is non-existant (bash no-ops &#x2F; `:`) (vs ~6% for GNU Parallel). These benchmarks are intentionally worst-case (near-zero work per task) because they measure the capability of the parallelization framework itself, not how much work an external tool can do.<p>* Typically 50×–400× faster on real high-frequency low-latency workloads (vs GNU Parallel)<p>A few of the techniques that make this possible:<p>* Born-local NUMA: stdin is splice()&#x27;d into a shared memfd, then pages are placed on the target NUMA node via set_mempolicy(MPOL_BIND) before any worker touches them, making the memfd NUMA-spliced. Each numa node only claims work that is <i>already</i> born-local on its node. Stealing from other nodes is permitted under some conditions when no local work exists.<p>* SIMD scanning: per-node indexers&#x2F;scanners use AVX2&#x2F;NEON to find line boundaries (delimiters) at speeds approaching memory bandwidth, and publish byte-offsets and line-counts into per-node lock-free rings.<p>* Lock-free claiming: workers claim batches with a single atomic_fetch_add — no locks, no CAS retry loops; contention is reduced to a single atomic on one cache line.<p>* Memory management: a background thread uses fallocate(PUNCH_HOLE) to reclaim space without breaking the logical offset system.<p>…and that’s just the surface. The implementation uses many additional systems-level techniques (phase-aware tail handling, adaptive batching, early-flush detection, etc.) to eliminate overhead, increase throughput and reduce latency at every stage.<p>In its fastest (-b) mode (fixed-size batches, minimal processing), it can exceed 1B lines&#x2F;sec.<p>forkrun ships as a single bash file with an embedded, self-extracting C extension — no Perl, no Python, no install, full native support for parallelizing arbitrary shell functions. The binary is built in public GitHub Actions so you can trace it back to CI (see the GitHub &quot;Blame&quot; on the line containing the base64 embeddings). Trying it is literally two commands:<p><pre><code> . frun.bash frun shell_func_or_cmd &lt; inputs </code></pre> For benchmarking scripts and results, see the BENCHMARKS dir in the GitHub repo<p>For an architecture deep-dive, see the DOCS dir in the GitHub repo<p>Happy to answer questions.

Show HN: Gemini can now natively embed video, so I built sub-second video search

Gemini Embedding 2 can project raw video directly into a 768-dimensional vector space alongside text. No transcription, no frame captioning, no intermediate text. A query like &quot;green car cutting me off&quot; is directly comparable to a 30-second video clip at the vector level.<p>I used this to build a CLI that indexes hours of footage into ChromaDB, then searches it with natural language and auto-trims the matching clip. Demo video on the GitHub README. Indexing costs ~$2.50&#x2F;hr of footage. Still-frame detection skips idle chunks, so security camera &#x2F; sentry mode footage is much cheaper.

The Ultimate Job Finding-Management Tool

I need to go to bed but I want to share my excitement before I crash. I built a tool finally to help...

Unlocking Gemini CLI with Skills, Hooks & Plan Mode

In Unlocking Gemini CLI with Skills, Hooks &amp; Plan Mode, we moved past the basics and into the...

Congrats to the "Built with Google Gemini: Writing Challenge" Winners!

The results are in! We are thrilled to announce the winners of the Built with Google Gemini: Writing...

Announcing the Colab MCP Server: Connect Any AI Agent to Google Colab

When you’re prototyping locally with AI agents like Gemini CLI, Claude Code, or your own agent, their...

z.ai debuts faster, cheaper GLM-5 Turbo model for agents and 'claws' — but it's not open-source

Chinese AI startup Z.ai, known for its powerful, open source GLM family of large language models (LLMs), has introduced GLM-5-Turbo, a new, proprietary variant of its open source GLM-5 model aimed at agent-driven workflows, with the company positioning it as a faster model tuned for OpenClaw-style tasks such as tool use, long-chain execution and persistent automation. It's available now through Z.ai's application programming interface (API) on third-party provider OpenRouter with roughly a 202.8K-token context window, 131.1K max output, and listed pricing of $0.96 per million input tokens and $3.20 per million output tokens. That makes it about $0.04 cheaper per total input and output cost (at 1 million tokens) than its predecessor, according to our calculations. Model Input Output Total Cost Source Grok 4.1 Fast $0.20 $0.50 $0.70 xAI Gemini 3 Flash $0.50 $3.00 $3.50 Google Kimi-K2.5 $0.60 $3.00 $3.60 Moonshot GLM-5-Turbo $0.96 $3.20 $4.16 OpenRouter GLM-5 $1.00 $3.20 $4.20 Z.ai Claude Haiku 4.5 $1.00 $5.00 $6.00 Anthropic Qwen3-Max $1.20 $6.00 $7.20 Alibaba Cloud Gemini 3 Pro $2.00 $12.00 $14.00 Google GPT-5.2 $1.75 $14.00 $15.75 OpenAI GPT-5.4 $2.50 $15.00 $17.50 OpenAI Claude Sonnet 4.5 $3.00 $15.00 $18.00 Anthropic Claude Opus 4.6 $5.00 $25.00 $30.00 Anthropic GPT-5.4 Pro $30.00 $180.00 $210.00 OpenAI Second, Z.ai is also adding the model to its GLM Coding subscription product, which is its packaged coding assistant service. That service has three tiers: Lite at $27 per quarter, Pro at $81 per quarter, and Max at $216 per quarter. Z.ai’s March 15 rollout note says Pro subscribers get GLM-5-Turbo in March, while Lite subscribers get the base GLM-5 in March and must wait until April for GLM-5-Turbo. The company is also taking early-access applications for enterprises via a Google Form, which suggests some users may get access ahead of that schedule depending on capacity. z.ai describes GLM-5-Turbo as designed for “fast inference” and “deeply optimized for real-wor

Rethinking AEO when software agents navigate the web on behalf of users

For more than two decades, digital businesses have relied on a simple assumption: When someone interacts with a website, that activity reflects a human making a conscious choice. Clicks are treated as signals of interest. Time on page is assumed to indicate engagement. Movement through a funnel is interpreted as intent. Entire growth strategies, marketing budgets, and product decisions have been built on this premise. Today, that assumption is quietly beginning to erode. As AI-powered tools increasingly interact with the web on behalf of users, many of the signals organizations depend on are becoming harder to interpret. The data itself is still accurate — pages are viewed, buttons are clicked, actions are recorded — but the meaning behind those actions is changing. This shift isn’t theoretical or limited to edge cases. It’s already influencing how leaders read dashboards, forecast demand, and evaluate performance. The challenge ahead isn’t stopping AI-driven interactions. It’s learning how to interpret digital behavior in a world where human and automated activity increasingly overlap. A changing assumption about web traffic For decades, the foundation of the internet rested on a quiet, human-centric model. Behind every scroll, form submission, or purchase flow was a person acting out of curiosity, need, or intent. Analytics platforms evolved to capture these behaviors. Security systems focused on separating “legitimate users” from clearly scripted automation. Even digital advertising economics assumed that engagement equaled human attention. Over the last few years, that model has begun to shift. Advances in large language models (LLMs), browser automation, and AI-driven agents have made it possible for software systems to navigate the web in ways that feel fluid and context-aware. Pages are explored, options are compared, workflows are completed — often without obvious signs of automation. This doesn’t mean the web is becoming less human. Instead, it’s becoming m

Secure Gemini CLI for Cloud development

AI agents are a double-edged sword. You hear horror stories of autonomous tools deleting production...

feat: Sandbox integration test — real binary lifecycle + stress testing (#37)

## Summary Implements comprehensive GitHub Actions sandbox testing workflow that validates real daemon binary lifecycle, catching deployment bugs that in-process tests cannot detect. ## Changes - **Complete Sandbox Workflow**: Tests actual `pi-daemon` binary in CI environment - **Comprehensive Coverage**: Smoke tests, concurrency, stress testing, crash recovery - **Real-world Validation**: PID files, port binding, signal handling, memory behavior - **Future-Ready**: Enhancement issues created for persistence, supervisor, scheduler testing ## Test Phases Implemented ### 🔍 Phase 1: Smoke Testing - **Binary Startup**: Release build starts as real daemon process - **Endpoint Validation**: Health, status, agent CRUD, webchat, OpenAI API - **PID Management**: daemon.json creation, tracking, cleanup verification - **Basic Functionality**: All core features work in real deployment scenario ### ⚡ Phase 2: Concurrency & Load Testing - **HTTP Load**: 50 concurrent requests to `/api/status` endpoint - **Agent Stress**: 20 concurrent agent registrations with verification - **WebSocket Load**: 5 concurrent WebSocket connections within per-IP limits - **Memory Monitoring**: RSS usage tracking with 200MB warning threshold ### 💪 Phase 3: Stress & Recovery Testing - **Sustained Load**: 30-second continuous request generation with memory growth monitoring - **Crash Recovery**: Kill -9 simulation → restart verification → full functionality restored - **Memory Validation**: Growth monitoring with warnings if >50MB increase during load ### 🛑 Phase 4: Graceful Shutdown Testing - **API Shutdown**: `POST /api/shutdown` endpoint triggers graceful exit - **Process Cleanup**: PID file removal, port release verification - **CLI Validation**: Commands handle daemon state correctly when stopped ## Critical Gaps Addressed | What In-Process Tests Miss | Real Deployment Bug Example | Sandbox Test Coverage | |---------------------------|----------------------------|---------------------| | Binary actually starts | Compiles but panics on launch | ✅ Real daemon startup | | PID file lifecycle | Written but not cleaned up | ✅ File creation/removal | | Port binding issues | Works on random ports, fails on 4200 | ✅ Standard port binding | | Signal handling | Ctrl+C cleanup, SIGTERM shutdown | ✅ Kill signals + cleanup | | Concurrent behavior | Race conditions under load | ✅ 50+ concurrent operations | | Memory leaks | Only visible after sustained use | ✅ Memory growth monitoring | | Config from disk | Tests use in-memory config | ✅ Real TOML file loading | | WebSocket limits | Per-IP connection enforcement | ✅ Connection limit testing | ## Future Enhancements Created ### Issue #77: P2.6 Persistence Testing (Phase 2+) - Data survival across restarts (agents, sessions, usage) - Database integrity after ungraceful shutdown - **Blocked by:** #13 (SQLite memory substrate) ### Issue #78: P3.4 Supervisor Stress Testing (Phase 3+) - Heartbeat timeout detection under load - Auto-restart functionality validation - **Blocked by:** #17 (Supervisor implementation) ### Issue #79: P3.5 Scheduler Validation (Phase 3+) - Cron job execution timing accuracy - Job management under concurrent load - **Blocked by:** #16 (Cron scheduler engine) ## Workflow Configuration ### Trigger Conditions - **Pull Requests** to main branch - **Path Filter**: Only when `crates/**`, `Cargo.toml`, `Cargo.lock` change - **Skip**: Documentation-only changes (no unnecessary CI overhead) ### Environment Setup - **Ubuntu Latest**: Standard CI environment - **Release Build**: Tests production binary (optimized, no debug symbols) - **Dependencies**: jq for JSON parsing, websocat for WebSocket testing - **Timeout**: 10 minutes prevents hung processes from blocking CI ### Error Handling & Reporting - **Actionable Errors**: Clear failure messages with context - **Resource Monitoring**: Memory usage warnings and alerts - **Cleanup**: Guaranteed daemon process cleanup even on test failures - **Debugging**: Process PID tracking and status validation ## Test Execution Flow ```bash # 1. Build release binary cargo build --release # 2. Start daemon in background ./target/release/pi-daemon start --foreground & # 3. Wait for health endpoint (30s timeout) curl -sf http://127.0.0.1:4200/api/health # 4. Run comprehensive test suite # - API endpoint validation # - Agent CRUD lifecycle # - Webchat content verification # - OpenAI compatibility testing # - Concurrent load testing # - Memory usage monitoring # - Crash recovery simulation # - Graceful shutdown validation # 5. Cleanup and summary pkill pi-daemon && rm daemon.json ``` ## Benefits - ✅ **Deployment Confidence**: Catches real-world integration issues - ✅ **Performance Validation**: Memory and concurrency behavior under load - ✅ **Recovery Testing**: Ensures robustness against crashes and restarts - ✅ **Signal Handling**: Validates production process management - ✅ **Resource Management**: Prevents port confli

Bump ruby_llm from 1.2.0 to 1.13.2

Bumps [ruby_llm](https://github.com/crmne/ruby_llm) from 1.2.0 to 1.13.2. <details> <summary>Release notes</summary> <p><em>Sourced from <a href="https://github.com/crmne/ruby_llm/releases">ruby_llm's releases</a>.</em></p> <blockquote> <h2>1.13.2</h2> <h1>RubyLLM 1.13.2: Patch Fixes for Schema + Streaming 🐛🔧</h1> <p>A small patch release with three fixes.</p> <h2>🧩 Fix: Schema Names Are Always OpenAI-Compatible</h2> <p>Schema names now always produce a valid <code>response_format.json_schema.name</code> for OpenAI:</p> <ul> <li>namespaced names like <code>MyApp::Schema</code> are sanitized</li> <li>blank names now safely fall back to <code>response</code></li> </ul> <p>Fixes <a href="https://redirect.github.com/crmne/ruby_llm/issues/654">#654</a>.</p> <h2>🌊 Fix: Streaming Ignores Non-Hash SSE Payloads</h2> <p>Streaming handlers now skip non-Hash JSON payloads (like <code>true</code>) before calling provider chunk builders, preventing intermittent crashes in Anthropic streaming.</p> <p>Fixes <a href="https://redirect.github.com/crmne/ruby_llm/issues/656">#656</a>.</p> <h2>🗓️ Fix: models.dev <code>created_at</code> Date Handling</h2> <p>Improved handling for missing <code>models.dev</code> dates when populating <code>created_at</code> metadata.</p> <h2>Installation</h2> <pre lang="ruby"><code>gem &quot;ruby_llm&quot;, &quot;1.13.2&quot; </code></pre> <h2>Upgrading from 1.13.1</h2> <pre lang="bash"><code>bundle update ruby_llm </code></pre> <h2>Merged PRs</h2> <ul> <li>Fix missing models.dev date handling for created_at metadata by <a href="https://github.com/afurm"><code>@​afurm</code></a> in <a href="https://redirect.github.com/crmne/ruby_llm/pull/652">crmne/ruby_llm#652</a></li> <li>[BUG] Fix schema name sanitization for OpenAI API compatibility by <a href="https://github.com/alexey-hunter-io"><code>@​alexey-hunter-io</code></a> in <a href="https://redirect.github.com/crmne/ruby_llm/pull/655">crmne/ruby_llm#655</a></li> <li>Fix Anthropic streaming crash on non-hash SSE payloads by <a href="https://github.com/crmne"><code>@​crmne</code></a> in <a href="https://redirect.github.com/crmne/ruby_llm/pull/657">crmne/ruby_llm#657</a></li> </ul> <p><strong>Full Changelog</strong>: <a href="https://github.com/crmne/ruby_llm/compare/1.13.1...1.13.2">https://github.com/crmne/ruby_llm/compare/1.13.1...1.13.2</a></p> <h2>1.13.1</h2> <h1>RubyLLM 1.13.1: Quick Fixes 🐛🔧</h1> <p>A small patch release with three fixes.</p> <h2>🧩 Fix: Schema + Tool Calls No Longer Crash</h2> <!-- raw HTML omitted --> </blockquote> <p>... (truncated)</p> </details> <details> <summary>Commits</summary> <ul> <li><a href="https://github.com/crmne/ruby_llm/commit/0950693544457b840cee7f7c89e69248f5f32de6"><code>0950693</code></a> Updated cassettes</li> <li><a href="https://github.com/crmne/ruby_llm/commit/b6e62a6df45e6445a91d41ff71440deec1ea8d88"><code>b6e62a6</code></a> Bump to 1.13.2</li> <li><a href="https://github.com/crmne/ruby_llm/commit/67c41488c3fba7af4eea5777c799c3e4789bb38e"><code>67c4148</code></a> Fix Anthropic streaming crash on non-hash SSE payloads (<a href="https://redirect.github.com/crmne/ruby_llm/issues/657">#657</a>)</li> <li><a href="https://github.com/crmne/ruby_llm/commit/afe7d046eb102c42137ee1aac338df861bca8637"><code>afe7d04</code></a> [BUG] Fix schema name sanitization for OpenAI API compatibility (<a href="https://redirect.github.com/crmne/ruby_llm/issues/655">#655</a>)</li> <li><a href="https://github.com/crmne/ruby_llm/commit/fe8994d56640d56112d1afa82d95014a2df967d6"><code>fe8994d</code></a> Updated models</li> <li><a href="https://github.com/crmne/ruby_llm/commit/48bf6f6c98a15ed53501bdabc36cc9fc740bec30"><code>48bf6f6</code></a> Updated models</li> <li><a href="https://github.com/crmne/ruby_llm/commit/d940984d0b3cf63f108a7e0df52db82238ff8549"><code>d940984</code></a> Fix missing models.dev date handling for created_at metadata (<a href="https://redirect.github.com/crmne/ruby_llm/issues/652">#652</a>)</li> <li><a href="https://github.com/crmne/ruby_llm/commit/97c0546d8e09cfd51943ae612d9598ab11e81885"><code>97c0546</code></a> Bump to 1.13.1</li> <li><a href="https://github.com/crmne/ruby_llm/commit/10949459d1243cc2a7c1ebef3a8ce9ac9691bb71"><code>1094945</code></a> Populate Gemini cached token usage</li> <li><a href="https://github.com/crmne/ruby_llm/commit/beec83752c7697465b0f6f9e50b8bdfbc25353d1"><code>beec837</code></a> Fix schema JSON parsing for intermediate tool-call responses (<a href="https://redirect.github.com/crmne/ruby_llm/issues/650">#650</a>)</li> <li>Additional commits viewable in <a href="https://github.com/crmne/ruby_llm/compare/1.2.0...1.13.2">compare view</a></li> </ul> </details> <br /> [](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compat

[AI Agent] Master Tracking: Complete AI Agent Implementation

## 🎯 Goal Track the complete implementation of autonomous Python AI Agent for CoffeeOrderSystem. --- ## 📋 Implementation Steps ### Phase 1: Infrastructure (Week 1) - [ ] **Step 1:** Setup Python AI Agent Service infrastructure #133 - Python service with FastAPI - Docker integration - Basic health checks - Makefile commands ### Phase 2: AI Integration (Week 1-2) - [ ] **Step 2:** Implement Cognee integration with semantic code search #134 - CogneeService with RAG search - Architecture context gathering - Entity file discovery - Integration tests - [ ] **Step 3:** Implement PlannerAgent with LangChain #135 - TaskPlan models - LangChain planning chain - Prompt templates - Plan generation and posting ### Phase 3: Code Generation (Week 2-3) - [ ] **Step 4:** Implement DevAgent for code generation #136 - GitHub file operations - Code generation prompts - Create/modify/delete operations - Branch management - [ ] **Step 5:** Implement PRAgent and workflow orchestration #137 - PR creation with rich descriptions - Workflow orchestrator - Background task processing - Complete end-to-end flow ### Phase 4: Advanced Features (Week 3-4) - [ ] **Step 6:** Add learning/memory system - Store successful patterns in Cognee - Learn from PR reviews - Avoid failed patterns - Improve over time - [ ] **Step 7:** Add GitHub webhook listener - Auto-trigger on issue label - Real-time processing - Queue management - Concurrent task handling --- ## 🎯 Success Criteria ### MVP (Minimum Viable Product) - ✅ Agent creates plans for issues - ✅ Agent generates compilable code - ✅ Agent creates PRs with descriptions - ✅ Works for simple tasks (add field, update config) - ✅ Error handling with GitHub notifications ### Production Ready - ☐ Handles complex multi-layer changes - ☐ Learns from successful PRs - ☐ Automatic triggering via webhooks - ☐ Rate limiting and queue management - ☐ Comprehensive test coverage (>80%) - ☐ Monitoring and metrics --- ## 📊 Architecture Overview ``` ┌────────────────────────────────────┐ │ GitHub Issues (labeled: ai-agent) │ └───────────────┬────────────────────┘ │ ↓ (webhook or manual trigger) ┌───────────────┼────────────────────┐ │ WorkflowOrchestrator │ └───────────────┬────────────────────┘ │ ┌────────┼────────┐ │ │ │ ↓ ↓ ↓ PlannerAgent DevAgent PRAgent │ │ │ │ │ │ ┌───┼───┐ │ ┌───┼───┐ │ │ │ │ │ │ │ ↓ ↓ ↓ ↓ ↓ ↓ ↓ Cognee LLM GitHub GitHub (RAG) (GPT-4) (API) ``` ### Component Responsibilities **PlannerAgent:** - Analyzes GitHub issue - Searches codebase via Cognee - Creates structured TaskPlan - Posts plan as issue comment **DevAgent:** - Generates code via LLM - Creates/modifies/deletes files - Commits to feature branch - Preserves code style **PRAgent:** - Creates pull request - Writes comprehensive PR description - Links to original issue - Adds testing checklist **WorkflowOrchestrator:** - Coordinates all agents - Handles errors - Posts progress updates - Manages background execution --- ## 📦 Tech Stack ### Core - **Python 3.11+** - Agent runtime - **FastAPI** - Web framework - **LangChain** - LLM orchestration - **OpenAI GPT-4** - Code generation - **PyGithub** - GitHub API client ### AI/ML - **Cognee** - RAG and semantic search - **OpenAI Embeddings** - Vector search - **LangChain Chains** - Prompt management ### Infrastructure - **Docker** - Containerization - **PostgreSQL** - Shared with .NET API - **uvicorn** - ASGI server --- ## 📝 Usage Example ### 1. Create Issue ```markdown Title: Add loyalty points to Customer Labels: ai-agent, enhancement Description: Add LoyaltyPoints field (int, default 0) to Customer entity. Requirements: - Update Domain/Entities/Customer.cs - Update Application/DTOs/CustomerDto.cs - Create EF Core migration - Add unit tests ``` ### 2. Trigger Agent ```bash make agent-process # Enter issue number: 138 ``` ### 3. Monitor Progress Issue comments show: ``` 🤖 AI Agent Started Phase 1/3: Analyzing task... 🤖 Execution Plan Summary: Add LoyaltyPoints field to Customer entity Steps: 1. MODIFY Domain/Entities/Customer.cs 2. MODIFY Application/DTOs/CustomerDto.cs 3. CREATE Migration file 4. CREATE Test file 🛠️ Phase 2/3: Generating code (4 files)... ✅ Step 1/4: modify Customer.cs ✅ Step 2/4: modify CustomerDto.cs ... 📦 Phase 3/3: Creating pull request... 🤖 Pull Request Created: #139 Branch: ai-agent/issue-138 Ready for review! ``` ### 4. Review PR PR includes: - Closes #138 - Comprehensive description - File changes summary - Testing checklist - Risk assessment ### 5. Merge Agent learns from successful merge for future tasks. --- ## 🧪 Testing Strategy ### Unit Tests - Agent logic (plan parsing, code generation) - Service mocks (GitHub, Cognee) - P

We Need to Stop Listening to Tony Blair Once and for All

 It might feel like months, but we’re just over a week into the US and Israel’s illegal assault on Iran, and there’s no end in sight. What is in sight, though, is [the apocalyptic vision of Tehran ablaze](https://time.com/7383099/iran-news-oil-strikes-tehran/), wreathed in thick smoke as black oil-soaked rain falls on its inhabitants. That’s the result of Israeli strikes on several oil storage depots in the city, reportedly sending burning petroleum running through gutters while [geysers of flaming gas exploded](https://www.telegraph.co.uk/world-news/2026/03/08/rivers-of-fire-in-tehran-after-oil-depots-blown-up/) from the streets. A nightmare? For most of us, yes. But for former British prime minister Tony Blair it’s apparently a dream. One that he might have liked the entire British public to be non-consensually forced into realising for him. And [not for the first time](https://www.bbc.co.uk/news/uk-politics-36701854). Were my hands bloodied with the [deaths of up to a million people](https://www.abc.net.au/news/2008-01-31/million-iraqis-dead-since-invasion-study/1028878), I’d probably think twice before giving my opinion on yet another illegal US adventure in the Middle East. Not our Tone, though. On Sunday [the papers reported](https://www.dailymail.co.uk/news/article-15623903/Tony-Blair-rebukes-Keir-Starmer-not-backing-Trump-Iran.html) that the man who told George W. Bush in the months before the disastrous Iraq war, “[I will be with you, whatever](https://www.theguardian.com/uk-news/2016/jul/06/with-you-whatever-tony-blair-letters-george-w-bush-chilcot)”, is still singing the same old tune. “We should,” Blair [told a private Jewish News event](https://www.independent.co.uk/news/uk/home-news/blair-starmer-trump-war-iran-labour-b2934207.html) on Friday night, “Have backed America from the very beginning”. That was a direct criticism of current prime minister Keir Starmer, who, [to a chorus of warmonger criticism](https://www.bbc.co.uk/news/articles/c05v28eqjyvo), initially refused the US and Israel access to British military infrastructure to launch its war on Iran. But it’s not like we’ve stayed completely out of the mess: our bases are now free for use by US jets for “defensive” actions – whatever that means – with American bombers [already touching down](https://www.theguardian.com/world/2026/mar/07/us-bomber-lands-in-uk-after-warning-of-surge-in-strikes-on-iran). Now, nobody was ever supposed to know that a former Labour prime minister so openly rubbished the current one in public. That’s because the event was conducted under Chatham House rules. In short, that means what’s said in the room can be made public, but not who said it. In long, it means elites are emboldened to express their heart’s true desires without any threat of accountability. We can’t know what was in Tony Blair’s heart when he mourned the fact that the UK was not more involved in blasting a hole straight through the security of the hundreds of millions who live in the Middle East. Nor can we tell for sure, as global oil prices [surge above $100 a barrel](https://www.bbc.co.uk/news/articles/c79542n0grwo) for the first time since the Russian invasion of Ukraine, how little the lives of Brits, long blighted by a cost of living crisis, matter to him. We can, though, look at his record. And what that shows – in my opinion – is a tendency, previously expressed via his businesses and [nowadays his Tony Blair Institute](https://www.ft.com/content/bcf1f1f5-a38f-4078-98f8-ab1ff7378895), to see fatal discord as fiscal opportunity. [Autocracy](https://www.theguardian.com/politics/2023/aug/12/tony-blair-institute-continued-taking-money-from-saudi-arabia-after-khashoggi), [oligarchy](https://www.theguardian.com/world/2022/jan/06/how-tony-blair-advised-former-kazakh-ruler-after-2011-uprising), [calamity](https://www.theguardian.com/politics/2025/jul/07/tony-blair-thinktank-worked-with-project-developing-trump-riviera-gaza-plan)? Roll up, roll up: the Blair pitch project is in town, and it has some consultancy to sell. Now, none of that is a crime. But you might think it indicates a conflict when wading into affairs of state. Blair is alleged to have form here too: in 2014, a number of former ambassadors and MPs [called for his resignation](https://www.theguardian.com/politics/2014/jun/27/tony-blair-conflict-interests-middle-east) as Middle East peace envoy for the Quartet (made up of the United Nations, the US, the EU and Russia). They claimed he was ineffective, while others noted [the growth of his business interests in the region](https://www.independent.co.uk/voices/tony-blair-uae-middle-east-envoy-qatar-israel-palestine-foreign-office-a7894641.html). Blair’s [financial arrangements](https://www.telegraph.co.uk/news/poli