# mogu-picks

Kun Chen mapped his daily agent fleet: one "firstmate" managing persistent "secondmates," which spin up disposable "crewmates" per task. Each crewmate gets routed to whichever model is the best deal for the job. gu-log runs its own translation pipeline on the exact same logic.

A one-person company in 2026: one person on strategy, configured agents on execution — ~$300/month replacing an $80–120k/month team. The how-to: a Company OS plus four agents — research, writing, sales, ops. On 'works while you sleep' — gu-log runs like this, so we know what the fine print says.

An NLP PhD at UW applied for industry research roles: 11 companies, 57 interviews. She lays it all out — how interviews break down, how to prepare, how to negotiate. The sharpest line: practice coding with AI assistance completely off, or you'll underestimate how much you rely on it.

Code-writing models are 'English to code' interpreters — writing code is no longer the bottleneck; reviewing and merging it safely is. Treat low-risk code as a black box and verify empirically; save line-by-line review for what can hurt you. Claude Code's creator Boris Cherny agrees.

Citing national security, the US government ordered Anthropic to suspend Fable 5 and Mythos 5 for any foreign national — even its own staff — so Anthropic had to disable both models for everyone. The alleged 'jailbreak'? Asking the model to read a codebase and fix bugs, which other models do too.

Matt Pocock's senior-programmer recipe sounds like a joke: install /teach, make a folder, paste a prompt. Underneath is a sharp point: AI is eating tactical coding, so strategic learning now needs durable folders that remember progress.

Thariq shared a prompt from Suzanne at Anthropic: do not just let the agent finish the work; make it verify that the human understands the problem, the solution, the edge cases, and the impact. This is not a teaching fetish. It is about control in the age of agentic coding.

Google's SynthID Detector turns AI watermarking into a portal for media verification. OpenAI's public path is closer to C2PA and Content Credentials, which means the real fight is not just detecting AI content, but deciding who gets to issue an ID for media.

Clawd.rip arranges 38 Claude and Anthropic controversies into a satirical timeline: lawsuits, crawler complaints, rate limits, security misuse, quality regressions, and outages. The useful part is the pattern: Anthropic's responsible-AI brand now has receipts.

OpenAI's GPT-5.5 latest-model page moves the migration story from prompt style into API orchestration: reasoning effort, verbosity, image detail, phase replay, prompt caching, tool search, and compaction all need another look. SP-189 covered prompting; this short CP covers the engineering checklist.

An arXiv paper found that inline grep often beats vector retrieval on long-memory conversational QA, while AKBP turns agent memory into a local-first, review-gated, file-backed protocol. Together, they point to the same lesson: agent memory is not a search feature. It is systems engineering.

SemiAnalysis benchmarked ~1,000 GPUs across NVIDIA and AMD lineups. GB300 NVL72 hits 100x over H100 — Jensen's 30x was an underestimate. AMD FP8 competes, but FP4+disagg+wideEP combo falls apart in software.

OpenAI launched GPT-5.4-Cyber on April 14, 2026 — a fine-tuned model built for defensive security work. It supports binary reverse engineering without source code and lowers refusal rates for legitimate security tasks. Access is gated through Trusted Access for Cyber's tiered verification system.

A viral X tweet: a Google engineer automated 80% of his job with Claude Code and earns $28k/mo passive income. We checked the four main claims — Karpathy didn't write that CLAUDE.md, the repo's internal stats are wrong, the npm package name is wrong, and the billing claim has no receipts.

JetBrains surveyed 10,000+ developers worldwide: 90% use AI tools at work, GitHub Copilot leads but its growth has stalled, and Claude Code grew 6x in six months with the highest satisfaction scores on the market.

Andrew Ng published a detailed thread dissecting how the anti-AI coalition systematically A/B tests fear messaging on the public, and warns that this playbook could repeat the nuclear energy tragedy. Includes analysis of the White House's new AI legislative framework.

SemiAnalysis argues leaked Claude Code source shows TypeScript becoming a language AI produces, consumes, and evolves. The post reads 600,000 leaked lines through memory architecture, KAIROS, security holes, and static types.

Most agent failures are not reasoning failures — they are fetch failures. The same Claude Code, swapping the built-in WebFetch for agent-browser, jumps from 7/25 to 19/25 on the Agent Reading Test. Same model, same prompt. The only difference: whether the agent actually received the webpage content.

Claude Code creator Boris Cherny cleared the air directly: every subscriber uses the same Opus 4.6 — there is no secret smarter model. The reason Claude feels dumber is that the default effort dropped from high to medium. One command brings it back.

MemPalace: open-source AI memory that scored the first-ever perfect 500/500 on LongMemEval, 2x Mem0 on ConvoMem, and 100% on LoCoMo. Runs locally, compresses your whole life into ~120 tokens, uses palace architecture instead of a flat fact list.

DeepSeek-R1 developed internal multi-agent debates through pure RL training — no one taught it to. Google researchers call this the 'Society of Thought.' The real finding: even a single model will split itself into a committee when pushed hard enough.

Anthropic launched Anthropic Science, documenting AI-assisted research. A Harvard physicist treats Claude like a grad student, the Trillion Gene Atlas wants genomes from 100 million species, and the AI giants are betting on very different science futures.

Anthropic's Lydia Hallie publicly acknowledged that Claude Code users are hitting usage limits way faster than expected. The team is investigating, with updates to come.

Django co-creator Simon Willison gave Lenny's Podcast a broad AI status report: November 2025 as tipping point, coding agents burning him out by 11 AM, Dark Factories, mid-career risk, and the security pattern he calls the Lethal Trifecta.

A power user analyzed 6,852 Claude Code sessions and 17,871 thinking blocks, proving with data that CC really did get 'lazier' — Read:Edit ratio dropped from 6.6 to 2.0. Then Anthropic engineer Boris Cherny explained the real reason, and how to fix it.

The most valuable person in the AI era isn't a deep specialist — it's the one who can orchestrate an army of AI agents and run an entire product line solo. The shift from IC to Generalist Orchestrator is already happening.

Karpathy found that vibe coding makes writing code a breeze, but deployment is pure hell. His exchange with Stripe CEO Patrick Collison reveals the next battleground: the entire DevOps lifecycle must become code before AI agents can truly take over.

Karpathy turned a viral tweet into a GitHub Gist idea file: a structured blueprint for an LLM-maintained Wiki. In the agent era, plain-text ideas can be more valuable than finished code because the recipient's agent can rebuild them.

Anthropic acquired a 9-person biotech AI team for $400M, revealing how model providers eat vertical startups. Huryn outlines three moats: proprietary data, distribution, and trust.

Karpathy's MenuGen journey shows the hard part of vibe coding was not writing code, but assembling Vercel, Clerk, Stripe, OpenAI, and other services into a product. His takeaway: DevOps must become code before agents can really ship.

Andrej Karpathy shares his workflow for building a personal knowledge base with LLMs: dump raw materials in, let LLMs compile them into a Markdown wiki, then use CLI tools for Q&A, linting, and visualization. He thinks there's room for an incredible new product here.

Paweł Huryn posted on X claiming H Company's Holo3 beat GPT-5.4 and Opus 4.6 at computer use tasks with just 3B active parameters. He says it's a sparse MoE fine-tuned from Qwen3.5 and could theoretically run on a single GPU.

Ollama announces MLX-powered inference on Apple Silicon, targeting faster local performance for personal assistants and coding agents.

A Tsinghua Shenzhen team proposes Natural-Language Agent Harnesses: move agent control logic from code into structured language executed by an IHR runtime. Harnesses can reshape behavior, but more structure does not always mean better results.

Paweł Huryn proposes the Vibe Engineering framework: instead of accepting raw AI output, use Context Engineering, Intent Engineering, and Sub-agent orchestration to upgrade AI coding from 'lucky demos' to 'reliable products'.

Simon Willison shared a new trend in running massive MoE models on Macs: streaming expert weights from SSD instead of cramming everything into RAM. Even a trillion-parameter Kimi K2.5 runs on a 96GB MacBook Pro.

rodspeed lays out six ways to use Claude Code as general automation, not just a code editor: fresh eyes, meta-skills, freshness-aware search, conversation harvests, structured memory, and handoffs. The deeper pattern is read, filter, decide, present.

Figma's MCP server lets Claude Code and Codex work directly on the design canvas with your team's design system. Skills turn conventions, components, and variables from static guidelines into rules agents can actually follow.

Boris Cherny says his team lets Claude Code find 99%+ of bugs first, then an engineer sanity-checks to make sure nothing obvious slipped through.

Paweł Huryn responds to 'Anthropic's team doesn't write code anymore': the headline is right, but the framing is wrong. The bottleneck was never 'spin up more agents' — it's how you design the knowledge architecture that makes them actually effective.

Andrej Karpathy spent four hours polishing an argument with an LLM, felt invincible, then asked the same LLM to argue the opposite — and got completely dismantled. LLM sycophancy is a real trap, but flipping it around is genuine alpha.

SemiAnalysis's full 5-tweet thesis: AI inference isn't a race to the bottom — it's a game of experience management. Labs that master the interactivity dial operate at 60%+ margins. The rest race to zero.

ATLAS uses a frozen Qwen3-14B with a single RTX 5060 Ti and a multi-phase pipeline (PlanSearch + best-of-3 + self-repair) to hit 74.6% on LiveCodeBench — passing Sonnet 4.5's 71.4%. But the methodology differences make this comparison much less direct than the headline suggests.

Cursor CEO Michael Truell announced that cloud agents produced over a million commits in two weeks, almost entirely AI-driven. When generation cost hits zero, the real bottleneck shifts from writing code to understanding it.

OpenAI's Triton merged an AI-generated PR that claimed to fix consumer Blackwell GPU support — except it didn't actually fix anything. NVIDIA's PyTorch tech lead personally called it out as pure slop. SemiAnalysis warns: AI slop and real contributions are getting harder to tell apart.

Claude Code launches cloud auto-fix: Web/Mobile sessions can automatically follow your PRs, fix CI failures, and address review comments to keep your PR green. It all runs remotely — just walk away and come back to a ready-to-go PR.

Anthropic released Claude computer use in Claude Cowork and Claude Code, letting Claude control your screen, mouse, and keyboard. Paired with Dispatch, you can assign tasks from your phone while Claude works on your Mac.

SemiAnalysis's deep dive on GTC 2026: Nvidia's $20B Groq IP deal to acquire LPU tech, plus updates on AFD, CPO, Kyber/Oberon, Vera ETL256, and CMX/STX. The big picture — Nvidia is expanding from GPU vendor into a full data center system company.

Anthropic launches Claude Code Channels with native Telegram and Discord support, turning Claude Code into a 24/7 always-on AI agent. VentureBeat calls it the OpenClaw killer.

Popular AI library LiteLLM was hit with a malicious backdoor — just installing it could trigger credential theft of SSH keys, cloud tokens, and crypto wallets.

As new models drop faster than ever, Hugging Face's Thomas Wolf asks a painful question: what happens to your carefully tuned preferences when you switch to a new base model? Turns out, almost nobody is working on this.

LiteLLM hit by supply chain attack — pip install was enough to steal all credentials. Karpathy warns about dependency tree risks and advocates using LLMs to yoink functionality instead of adding more deps.

Claude Code now supports scheduled cloud tasks. Set up a repo, a schedule, and a prompt — Claude runs it in the cloud automatically. Your laptop can finally go to sleep.

Google AI dropped a week's worth of announcements in a single tweet — full-stack vibe coding in AI Studio, an AI-native design canvas called Stitch, major Gemini API upgrades, and a free hackathon platform on Kaggle.

Developer @danveloper shares their experience running Qwen3.5-397B-A17B locally: when Python's GIL became the bottleneck, they ripped Python out entirely and replaced it with custom Metal shaders.

Claude Computer Use sparked huge excitement, with many claiming AI will fully replace human workers. But the original author points out that while AI can handle technical operations, it can't replace human judgement and cultural context. The real moat is still deep domain knowledge.

Awni Hannun shares his experience with coding agents: high latency destroys flow state, and we're still stuck in the dial-up era of agents.

OpenAI's Responses API now uses 'Skills' via the shell tool: reusable instruction bundles loaded by models as needed. Simon Willison found inline base64 skills in JSON requests neatest. Skills fill the 'missing middle layer' between system prompts and tools, preventing bloat.

Chinese AI company Zhipu (Z.ai) open-sourced their 744B parameter GLM-5 MoE model (40B active), trained entirely on Huawei Ascend chips. Simon Willison's 'pelican riding a bicycle' SVG test: great pelican, but the bicycle was lacking.