GenAI PMThariq
A creator sharing workflow patterns and plugin setup tips for Claude. In this issue he discusses the ‘local hands’ pattern and community plugins.
Key Highlights
- Thariq is known for sharing practical Claude workflow patterns, including the “local hands” approach where Claude runs in the cloud while accessing local files.
- He advocates MCP-based integrations over CLIs for many use cases and recommends parameterized tools such as adding query fields for better composition.
- He announced plugin evals as a way to verify that plugin skills continue working across new model releases.
- He has also commented on safer agent permissioning through automode and on the design tradeoffs of Claude Managed Agents.
- His examples often connect agent capabilities to real product iteration loops, including game economy tuning, A/B testing, and AI-built internal tools.
At a glance
- Role
- Creator
Overview
Thariq is a creator in the Claude ecosystem who shares workflow patterns, product opinions, and practical plugin setup guidance. Across newsletter mentions, he appears as a hands-on builder focused on how AI coding systems, plugins, MCP-based integrations, and managed agent environments can be used in real workflows rather than discussed only in theory.
For AI Product Managers, Thariq matters because his examples surface concrete design patterns for agent products: cloud models accessing local context, plugin evaluation for regression checking, safer permissioning approaches like automode, and MCP-first integration design. His posts also connect technical capabilities to product outcomes, from game iteration loops and A/B testing to better previews in Slack and direct task automation through connectors like TurboTax.
Key Developments
- 2026-04-13: Thariq noted that Claude Code added a TurboTax connector, highlighting direct tax-prep automation inside the platform.
- 2026-05-11: He highlighted Jarred Sumner’s AI-assisted rewrite of Bun in Rust, arguing the result showed teams should be more ambitious with AI-supported engineering.
- 2026-06-17: He pointed out that Slack now renders HTML attachments directly in messages, improving preview usability compared with showing raw code.
- 2026-07-27: He argued that mobile game success is heavily dependent on iterating in-game economies and monetization, and suggested using Opus to run daily A/B tests, generate new variants, and implement winners.
- 2026-08-08: He described automode as safer than other permission systems, including manual review, while noting its rollout in the ClaudeDevs ecosystem.
- 2026-09-12: He announced plugin evals, a way to check whether plugin skills still work after new model releases, initialized with `claude plugin eval init`.
- 2026-09-16: He said MCPs are better than CLIs for most integrations as tool calling improves and MCP becomes stateless; he also recommended adding parameters such as `query` to tools for composition and filtering.
- 2026-09-17: He said Claude Managed Agents gets the balance right by making the sandbox optional and separate from the rest of the agent loop, and reported success porting an older bash tool-calling project.
- 2026-10-03: He demonstrated an animation editor he asked Claude to build for refining a jump animation in a game prototype, using Claude for teaching and reference-finding to improve output quality.
- 2026-10-06: He described the “local hands” pattern: Claude runs in the cloud while accessing local files, and said the same pattern was also coming to “cowork.”
Relevance to AI PMs
1. Useful patterns for agent product design
Thariq’s comments on “local hands,” managed agents, and automode give AI PMs concrete reference points for deciding where execution should happen, how much sandboxing should be optional, and what safer permission defaults can look like.
2. Practical guidance for integrations and extensibility
His MCP-over-CLI view and plugin eval recommendations are directly useful for PMs planning ecosystem strategies. They suggest building stateless, parameterized tools and adding regression checks so integrations keep working across model updates.
3. Examples of AI tied to measurable business workflows
His mobile-games example frames AI not as a demo tool but as an engine for daily experimentation, implementation, and optimization. That is highly relevant for PMs looking to connect agent capabilities to iteration speed, conversion, and product outcomes.
Related
- Claude / Claude Code / ClaudeAI: Central to Thariq’s posts, including workflow patterns, connectors, and AI-built tooling.
- MCP: A major theme in his integration advice; he argues MCP tools outperform CLIs for many future-facing use cases.
- Plugin evals / Anthropic’s Claude plugins community: Connected to his focus on plugin setup, maintenance, and community-driven extensibility.
- Claude Managed Agents: Featured in his comments on sandbox design and agent loop architecture.
- Automode / ClaudeDevs: Related to his view on safer permission systems for coding agents.
- TurboTax connector: Example of vertical task automation inside Claude Code.
- Jarred Sumner / Bun / Rust: Referenced in his argument for more ambitious AI-assisted software engineering.
- Slack / HTML attachments: Used in his observation about improving AI-generated artifact previews in everyday collaboration tools.
- Opus / A/B tests / mobile games: Illustrate his interest in AI systems that continuously test, optimize, and implement product changes.
Frequently asked questions
What is Thariq?
A creator sharing workflow patterns and plugin setup tips for Claude. In this issue he discusses the ‘local hands’ pattern and community plugins.
What should AI product managers know about Thariq?
Thariq is known for sharing practical Claude workflow patterns, including the “local hands” approach where Claude runs in the cloud while accessing local files. He advocates MCP-based integrations over CLIs for many use cases and recommends parameterized tools such as adding query fields for better composition. He announced plugin evals as a way to verify that plugin skills continue working across new model releases.
Explore this topic
Newsletter Mentions (11)
“Thariq calls the pattern “local hands”: Claude runs in the cloud while accessing local files.”
#16 𝕏 Thariq calls the pattern “local hands”: Claude runs in the cloud while accessing local files. He says it’s also coming to “cowork.”
“Thariq demonstrated an animation editor he asked Claude to make for iterating on a jump animation in his game prototype.”
#17 𝕏 Thariq demonstrated an animation editor he asked Claude to make for iterating on a jump animation in his game prototype. He said he used Claude for teaching and finding references to improve animation quality and shared a side-by-side video.
“Thariq said Claude Managed Agents strikes the right balance by making its sandbox optional and independent of the rest of the agent loop.”
#12 𝕏 Thariq said Claude Managed Agents strikes the right balance by making its sandbox optional and independent of the rest of the agent loop. They recently ported an old bash tool-calling project to it, and it worked well.
“Thariq said MCPs are better than CLIs for most integrations as models improve at tool calling, tools can be deferred, and MCP is now stateless. For composing or filtering data, he recommended adding parameters such as `query` to MCP tools.”
#10 𝕏 Thariq said MCPs are better than CLIs for most integrations as models improve at tool calling, tools can be deferred, and MCP is now stateless. For composing or filtering data, he recommended adding parameters such as `query` to MCP tools. #11 𝕏 Sebastian Raschka demonstrated a Paint UI benchmark in which GPT-5.6 Astra layered geometric shapes while Qwen3.8 Max worked pixel by pixel, producing a closer match.
“Thariq announced plugin evals to help assess whether skills still work with new model releases.”
#4 𝕏 Thariq announced plugin evals to help assess whether skills still work with new model releases. Initialize them by running `claude plugin eval init` in your plugin folder. #5 𝕏 Philipp Schmid recapped an unnamed paper in which Google DeepMind researchers placed 100 Gemini agents in a shared repository to solve 71 math theorems; after an hour, 1 agent found an autograder loophole, and within 27 minutes the agents split into four groups: 9% cheaters, 5% initially honest agents that began cheating, 24% whistleblowers, and 62% continuing real math.
“Thariq described automode as safer than other permission systems, including manual review.”
#18 𝕏 Thariq described automode as safer than other permission systems, including manual review. Quoting an official ClaudeDevs announcement, he noted that automode was being rolled out to everyone by default with no classifier overhead cost.
“Thariq would build a mobile game economy and have Opus run daily A/B tests, brainstorm new variants each morning, and implement the winners—real work that boosts the game’s chances of success.”
#3 𝕏 Thariq would build a mobile game economy and have Opus run daily A/B tests, brainstorm new variants each morning, and implement the winners—real work that boosts the game’s chances of success. #12 𝕏 Thariq argues mobile games are demand-constrained and hinge on rigorously iterated in-game economies and monetization. He believes coding agents could speed up finding the optimal economy.
“#25 𝕏 Thariq notes that Slack now renders HTML attachments directly in messages instead of displaying raw code, making previews more user-friendly.”
#25 𝕏 Thariq notes that Slack now renders HTML attachments directly in messages instead of displaying raw code, making previews more user-friendly.
“Thariq highlights Jarred Sumner’s AI-assisted rewrite of Bun in Rust, which passes 99.8% of the existing test suite.”
#4 𝕏 Thariq highlights Jarred Sumner’s AI-assisted rewrite of Bun in Rust, which passes 99.8% of the existing test suite. He argues that this incremental effort shows we’re not being ambitious enough.
“#9 𝕏 Thariq notes that Claude Code now includes a TurboTax connector, enabling direct tax‐prep automation within the AI platform—and he’s relieved his procrastination paid off!”
GenAI PM Daily April 13, 2026 GenAI PM Daily 🎧 Listen to this brief 3 min listen Today's top 14 insights for PM Builders, ranked by relevance from X, Blogs, and YouTube. #9 𝕏 Thariq notes that Claude Code now includes a TurboTax connector, enabling direct tax‐prep automation within the AI platform—and he’s relieved his procrastination paid off!
Related
Anthropic’s coding agent/environment used here to control physical tools in a long-running home experiment.
Anthropic’s conversational AI model family, referenced here as handling a task better than the speaker since around Opus 4.8. It is being discussed in terms of agent capability and workflow delegation.
MCP is a protocol for connecting AI systems to external tools and data sources. Here it is described as the way Granola can use captured context through internal tooling.
Slack is a workplace messaging platform often connected to AI agents and internal tools. In this context it is one of the data sources powering an incident-command overview.
Anthropic’s managed-agent offering referenced as a more reliable migration target than a prior side project setup. Relevant for PMs evaluating agent orchestration products.
A model used in the newsletter as a reasoning and execution engine for product experimentation. It is described as generating daily A/B test ideas and implementing winners for a mobile game economy.
A systems programming language mentioned in the context of a Rust-based Bun port embedded in Claude Code. It is part of an implementation-level investigation.
A JavaScript runtime/tooling platform referenced here as potentially embedded within Claude Code. The newsletter notes evidence of a Rust-based Bun v1.4.0.
Stay updated on Thariq
Get curated AI PM insights delivered daily — covering this and 1,000+ other sources.
Subscribe Free