aisettings
Log InAdd your setup

Thariq Shihipar

Works at a lab
Last updated: Sep 3, 2026
RoleClaude Code engineer, Anthropic
EarnsWorks at a lab
Sources13 cited fields · high confidence
@trq212www.thariq.io

Uses Claude Code with visual planning artifacts, voice prompts, verifiable goals, and parallel agent work.

“Plans should optimize for human review, not just model readability.” chatprd.ai

Share cardShare your setupCompare

Loadout

aisettingsShare on X

Thariq Shihipar’s loadout

Works at a lab
Harness
Claude Code
Thariq continues to describe and develop Claude Code, including planned function hooks and greater hackability.
Model
Claude Fable 5.1
lower for low-verification tasks
Also uses
Claude Opus 5
Thariq recommends Opus 5 as a daily driver paired with Fable for planning, brainstorming, and hardest bugs.
ToolsCLAUDE.md and skills+2 more

3 views · 0 equipped

Models

Claude Fable 5.1lower for low-verification tasks
His Sep 2 post responds to the Fable 5.1 launch and calls it a very good model.
x.com, 2026-09 ↗
Claude Opus 5
Thariq recommends Opus 5 as a daily driver paired with Fable for planning, brainstorming, and hardest bugs.
x.com, 2026-07 ↗

Tools

CLAUDE.md and skills
Thariq's context-engineering work focuses on reducing and right-sizing system prompts, skills, and CLAUDE.md files for newer models.
x.com, 2026-07 ↗
Wispr Flow
Dictates Claude Code prompts rather than typing them.
creatoreconomy.so, 2026-07 ↗
Slack agent orchestration
Runs a team of agents on different tasks in parallel.
creatoreconomy.so, 2026-07 ↗

Workflow

  1. 1Use lower effort for tasks needing less verification or with fewer edge cases; reserve higher effort for harder work.x.com, 2026-09
  2. 2Pair Opus 5 as a daily driver with Fable for planning, brainstorming, and the hardest bugs.x.com, 2026-07
  3. 3Strip unnecessary prompt scaffolding for frontier models and tailor system prompts per model.x.com, 2026-07
  4. 4Ask Claude to brainstorm and plan in interactive HTML instead of giant Markdown documents.chatprd.ai, 2026-05
  5. 5Use disposable micro-apps to edit difficult decisions, then export the result back into the plan.chatprd.ai, 2026-05
  6. 6Read the plan manually before execution and use /goal only with a verifiable endpoint.creatoreconomy.so, 2026-07

Timeline

  1. 2026-09-03HarnessClaude Codex.com, 2026-09
  2. 2026-09-02ModelClaude Fable 5.1x.com, 2026-09
  3. 2026-09-02Effortlower for low-verification tasksx.com, 2026-09
  4. 2026-07-24Also usesClaude Opus 5x.com, 2026-07
  5. 2026-07-24ToolsCLAUDE.md and skillsx.com, 2026-07
  6. 2026-07-19ToolsWispr Flowcreatoreconomy.so, 2026-07
  7. 2026-07-19ToolsSlack agent orchestrationcreatoreconomy.so, 2026-07

Comments

Log in to comment