Super User Daily: April 08, 2026
April 6 was dominated by the OpenClaw ban aftermath and its ripple effects across the Claude Code ecosystem. Users are scrambling to rewire their agent stacks, discovering new multi-agent architectures in the process, and a few builders quietly shipped tools that could matter more than the drama itself.
@ens_pyrz [Claude Code]
https://x.com/ens_pyrz/status/2041251711444976080
Walked outside, brainstormed an app idea using Claude chat, came back, dropped a TODO.md into Claude Code with Firebase CLI access, and had a working app deployed within an hour. The workflow of separating ideation from execution and letting Claude Code handle the build with a pre-written spec is exactly how non-engineers should be thinking about shipping.
@09pauai [Claude Code]
https://x.com/09pauai/status/2041108763063849433
Built a market research video in 10 minutes by combining Claude Code, Remotion, and an affiliate monetization layer. The stack is Claude Code for market research, Remotion for automated video production, and affiliate links for revenue. Claims this workflow can generate 100K yen per month. Video creation difficulty is collapsing.
@kirubaakaran [Claude Code]
https://x.com/kirubaakaran/status/2041155485680931024
Tried to fetch commodity expiry dates from MCX (Indian stock exchange), but MCX blocks automated requests. Claude Code automatically figured out a workaround and delivered the data. Also tested with OpenClaw but Claude Code was faster. This is the kind of non-obvious automation where AI agents shine: navigating anti-bot protections without being told to.
@dgalarza [Claude Code]
https://x.com/dgalarza/status/2041169301923643410
Ran autoresearch as a Claude Code skill: 65 iterations across three rounds, fully autonomous. The loop is dead simple: make a change, commit, check if the metric improved, run a quality guard, keep or roll back, repeat. This is the autoresearch pattern applied to real production code improvement, not a toy experiment.
@akhsaas [Claude Code]
https://x.com/akhsaas/status/2041127028146196815
Running a multi-agent architecture: OpenClaw with Minimax M2.7 as the orchestrator, which delegates coding tasks to Claude Code running Opus and Sonnet underneath. This layered approach uses a cheap model for orchestration and expensive models only for the hard coding work. The $10/month Minimax plan makes this surprisingly affordable.
@ByanJati [Claude Code]
https://x.com/ByanJati/status/2041303373207810484
Set up Honcho memory to share context between Hermes Agent and Claude Code via Docker. The idea is that when Claude Code writes to shared memory, Hermes can read it and maintain continuity across agent handoffs. Setup requires separate embedding model and Gemini/OpenAI key. Still testing whether the closed loop actually works in practice.
@0thernet [Claude Code]
https://x.com/0thernet/status/2041252637723574368
Zo is a cloud computer where you can run Claude Code, Codex, or Gemini in an always-on terminal. Uses ACP protocol by Zed to stream output from Claude Code running locally. After the OpenClaw ban, Zo positions itself as a clean alternative since it does not intercept OAuth tokens. You can also run OpenClaw inside Zo.
@its_kuroo_sol [Claude Code]
https://x.com/its_kuroo_sol/status/2041211957898822115
An AI job search system built with Claude Code that submitted 700+ applications and actually landed someone a job. Hit 8K stars on GitHub in a single day. The system handles resume tailoring, application submission, and follow-up tracking. Claude Code as a personal job-hunting agent is one of the more practical non-coding applications.
@JuribaSaleh [Claude Code]
https://x.com/JuribaSaleh/status/2040999587696038392
Built a plugin where Claude Code handles implementation while Codex acts as the code reviewer. Combined with ralph loop, you get a dual-agent workflow for confident code changes. One AI writes, another AI reviews. The adversarial setup catches more bugs than either alone.
@gasikaramada [Claude Code]
https://x.com/gasikaramada/status/2041054970489737451
Created a marketplace-dev skill that converts Claude Code skills into installable plugins for a standard plugin marketplace. Install it with one command and your skills become distributable packages. This is infrastructure work that could matter a lot as the skills ecosystem grows.
@klemens_gc [Claude Code]
https://x.com/klemens_gc/status/2041175768907039208
Built a CompanyOS using Claude Code and Obsidian as the knowledge base stack. The idea is treating your entire knowledge management system as a Claude Code workspace. Shared the template publicly.
@ampersend_ai [OpenClaw]
https://x.com/ampersend_ai/status/2041293231443583046
Published a guide for giving your OpenClaw agent a crypto wallet in under 5 minutes. Practical DeFi integration that turns OpenClaw into an agent that can actually transact on-chain.
@Stef4h18 [Claude Code]
https://x.com/Stef4h18/status/2041068780780986821
A project that would have required over 60,000 euros in traditional development costs was built with Claude Code for 200 euros total investment. Even accounting for exaggeration, the 300x cost reduction claim reflects a real pattern in the data.
@john_v_shrader [Claude Code]
https://x.com/john_v_shrader/status/2040984324196204754
Using Claude Code hooks with before_tool_call in settings.json to automatically check repo state before every edit. Saves context tokens by preventing Claude from working on stale state. Hooks are underutilized infrastructure.
@hillsidedev_ [Claude Code]
https://x.com/hillsidedev_/status/2040964702625509729
JetBrains surveyed 10,000 developers. Claude Code went from 3% to 18% adoption in under a year. More interesting: 91% satisfaction score and NPS of 54. Copilot has the install base, Claude Code has the loyalty. There is a difference between the tool that comes pre-installed and the tool you go out of your way to get.
@CryptoBandit [OpenClaw]
https://x.com/CryptoBandit/status/2041032542304817537
Using OpenClaw primarily for music production workflows and promotional content. Non-coding creative application that demonstrates agent utility beyond the developer bubble.
@EvanHoffman [Claude Code]
https://x.com/EvanHoffman/status/2041150951449309599
Discovered that asking Claude Code to review a 130-line CLAUDE.md file on the Pro plan consumed 14% of session usage. This highlights a real optimization problem: your configuration file itself is eating into your work budget.
@getcontexto [OpenClaw]
https://x.com/getcontexto/status/2041081620883701920
Contributed a fix to OpenClaw core that makes the context engine handle context persistently. Before this fix, context was being dropped between sessions. Open-source contribution that improves the fundamental agent experience.
🗣 User Voice
User Voice
Aggressive token consumption is the number one pain point right now. @Bartek95058577 ran analysis showing x66 less usage than a week ago on the x5max plan, with 40% of 12 million tokens burned on just two research sessions. @adamsbytes burned through the MAX 20x plan in 48 hours of programming only, actively discouraging subagent usage. @EvanHoffman found that even reviewing CLAUDE.md eats 14% of a Pro session. The theme is clear: power users are hitting walls faster than ever.
Platform lock-in fear is growing. @NoahHirshon built an entire system around Claude Code but is now scared of lock-in after recent pricing changes. @BhagyaMudgal tried switching to Codex but found it too painful to replicate the Claude Code setup built over a year. The switching cost is real and Anthropic knows it.
OpenClaw users are frustrated with post-ban quality. @0daycore reports GPT works terribly in OpenClaw with constant forgetting and endless message loops. @Stocko_Steve and others are begging for Claude to come back. The forced migration to alternative models is exposing quality gaps.
Security researchers caught in the crossfire. @PhiloGroves reports Claude Code flagging legitimate security research as usage violations despite not using OpenClaw. The enforcement net is wider than intended.
Aggressive token consumption is the number one pain point right now. @Bartek95058577 ran analysis showing x66 less usage than a week ago on the x5max plan, with 40% of 12 million tokens burned on just two research sessions. @adamsbytes burned through the MAX 20x plan in 48 hours of programming only, actively discouraging subagent usage. @EvanHoffman found that even reviewing CLAUDE.md eats 14% of a Pro session. The theme is clear: power users are hitting walls faster than ever.
Platform lock-in fear is growing. @NoahHirshon built an entire system around Claude Code but is now scared of lock-in after recent pricing changes. @BhagyaMudgal tried switching to Codex but found it too painful to replicate the Claude Code setup built over a year. The switching cost is real and Anthropic knows it.
OpenClaw users are frustrated with post-ban quality. @0daycore reports GPT works terribly in OpenClaw with constant forgetting and endless message loops. @Stocko_Steve and others are begging for Claude to come back. The forced migration to alternative models is exposing quality gaps.
Security researchers caught in the crossfire. @PhiloGroves reports Claude Code flagging legitimate security research as usage violations despite not using OpenClaw. The enforcement net is wider than intended.
📡 Eco Products Radar
Eco Products Radar
Minimax M2.7 — Budget orchestration model gaining traction with OpenClaw users, $10/month plan
OpenClaw — Open-source AI agent, navigating post-Claude-ban era with GPT and other providers
Codex/OpenAI — Increasingly used as complementary tool alongside Claude Code
Hermes Agent — NousResearch alternative agent framework
Gemma 4 — Google open-source model, running locally via Ollama
MCP (Model Context Protocol) — Standard protocol for extending Claude Code capabilities
Remotion — Programmatic video creation framework
Honcho — Cross-agent memory system
ACP (Agent Communication Protocol) — Zed protocol for streaming agent output
Minimax M2.7 — Budget orchestration model gaining traction with OpenClaw users, $10/month plan
OpenClaw — Open-source AI agent, navigating post-Claude-ban era with GPT and other providers
Codex/OpenAI — Increasingly used as complementary tool alongside Claude Code
Hermes Agent — NousResearch alternative agent framework
Gemma 4 — Google open-source model, running locally via Ollama
MCP (Model Context Protocol) — Standard protocol for extending Claude Code capabilities
Remotion — Programmatic video creation framework
Honcho — Cross-agent memory system
ACP (Agent Communication Protocol) — Zed protocol for streaming agent output
Comments