Keep on to blur preview images; turn off to show them clearly
RT @kirancodes: @whynotqat @SebastienBubeck the proof is accessible online here and looks fairly readable: https://t.co/RMXwsoVkmH the mai…
Root node of the web of threads: https://t.co/ifH80GcLpo
![I installed this morph warp search tool this morning and then sort of forgot about it. I wondered if my agents had been using it and if it was helping them at all.
So I asked them and they all admitted to a man (err, robot) that they hadn’t used it at all and had instead stuck to using grep as is their wont. Already like grizzled graybeard devs stuck in their ways.
Anyway, I asked them to poke around and try it and let me know what they thought. So they took it for a test drive and had some feedback.
Then I asked them to figure out when and why and how they could best use it, and then had one of them come up with the most token-dense, intuitive explanation (at least intuitive to their robot brain) to explain all that, for use in an AGENTS dot md file.
Here’s what it came up with; feel free to steal it for your own AGENTS dot md (and btw, my verdict for now is that it’s worth keeping around for a while to see how much they end up using it, but I think it’s probably pretty useful-- if they can actually be convinced to use it, which is why these little blurbs are so critical, and why they should focus on convincing rather than ordering the agents! Moral suasion ain’t just for humans…):
---
### Morph Warp Grep vs Standard Grep
Warp Grep = AI agent that greps, reads, follows connections, returns synthesized context with line numbers.
Standard Grep = Fast regex match, you interpret results.
Decision: Can you write the grep pattern?
- Yes → Grep
- No, you have a question → mcp__morph-mcp__warp_grep
#### Warp Grep Queries (natural language, unknown location)
"How does the moderation appeals flow work?"
"Where are websocket connections managed?"
"What happens when a user submits a post?"
"Where is rate limiting implemented?"
"How does the auth session get validated on API routes?"
"What services touch the moderationDecisions table?"
#### Standard Grep Queries (known pattern, specific target)
pattern="fileAppeal" # known function name
pattern="class.*Service" # structural pattern
pattern="TODO|FIXME|HACK" # markers
pattern="processenv" path="apps/web" # specific string
pattern="import.*from [']@/lib/db" # import tracing
#### What Warp Grep Does Internally
One query → 15-30 operations: greps multiple patterns → reads relevant sections → follows imports/references → returns focused line ranges (e.g., l3-guardian.ts:269-440) not whole files.
#### Anti-patterns
| Don't Use Warp Grep For | Why | Use Instead |
|------------------------|-----|-------------|
| "Find function handleSubmit" | Known name | Grep pattern="handleSubmit" |
| "Read the auth config" | Known file | Read file_path="lib/auth/..." |
| "Check if X exists" | Boolean answer | Grep + check results |
| Quick lookups mid-task | 5-10s latency | Grep is 100ms |
#### When Warp Grep Wins
- Tracing data flow across files (API → service → schema → types)
- Understanding unfamiliar subsystems before modifying
- Answering "how" questions that span 3+ files
- Finding all touching points for a cross-cutting concern
—
You can see the whole short interaction in this gist:
https://t.co/Z6jvFXDxub](/_next/image?url=https%3A%2F%2Fpbs.twimg.com%2Fprofile_images%2F1225476100547063809%2F53jSWs7z_400x400.jpg&w=3840&q=75 "I installed this morph warp search tool this morning and then sort of forgot about it. I wondered if my agents had been using it and if it was helping them at all.
So I asked them and they all admitted to a man (err, robot) that they hadn’t used it at all and had instead stuck to using grep as is their wont. Already like grizzled graybeard devs stuck in their ways.
Anyway, I asked them to poke around and try it and let me know what they thought. So they took it for a test drive and had some feedback.
Then I asked them to figure out when and why and how they could best use it, and then had one of them come up with the most token-dense, intuitive explanation (at least intuitive to their robot brain) to explain all that, for use in an AGENTS dot md file.
Here’s what it came up with; feel free to steal it for your own AGENTS dot md (and btw, my verdict for now is that it’s worth keeping around for a while to see how much they end up using it, but I think it’s probably pretty useful-- if they can actually be convinced to use it, which is why these little blurbs are so critical, and why they should focus on convincing rather than ordering the agents! Moral suasion ain’t just for humans…):
---
### Morph Warp Grep vs Standard Grep
Warp Grep = AI agent that greps, reads, follows connections, returns synthesized context with line numbers.
Standard Grep = Fast regex match, you interpret results.
Decision: Can you write the grep pattern?
- Yes → Grep
- No, you have a question → mcp__morph-mcp__warp_grep
#### Warp Grep Queries (natural language, unknown location)
\"How does the moderation appeals flow work?\"
\"Where are websocket connections managed?\"
\"What happens when a user submits a post?\"
\"Where is rate limiting implemented?\"
\"How does the auth session get validated on API routes?\"
\"What services touch the moderationDecisions table?\"
#### Standard Grep Queries (known pattern, specific target)
pattern=\"fileAppeal\" # known function name
pattern=\"class.*Service\" # structural pattern
pattern=\"TODO|FIXME|HACK\" # markers
pattern=\"processenv\" path=\"apps/web\" # specific string
pattern=\"import.*from [']@/lib/db\" # import tracing
#### What Warp Grep Does Internally
One query → 15-30 operations: greps multiple patterns → reads relevant sections → follows imports/references → returns focused line ranges (e.g., l3-guardian.ts:269-440) not whole files.
#### Anti-patterns
| Don't Use Warp Grep For | Why | Use Instead |
|------------------------|-----|-------------|
| \"Find function handleSubmit\" | Known name | Grep pattern=\"handleSubmit\" |
| \"Read the auth config\" | Known file | Read file_path=\"lib/auth/...\" |
| \"Check if X exists\" | Boolean answer | Grep + check results |
| Quick lookups mid-task | 5-10s latency | Grep is 100ms |
#### When Warp Grep Wins
- Tracing data flow across files (API → service → schema → types)
- Understanding unfamiliar subsystems before modifying
- Answering \"how\" questions that span 3+ files
- Finding all touching points for a cross-cutting concern
—
You can see the whole short interaction in this gist:
https://t.co/Z6jvFXDxub")
I installed this morph warp search tool this morning and then sort of forgot about it. I wondered if my agents had been using it and if it was helping them at all. So I asked them and they all admitted to a man (err, robot) that they hadn’t used it at all and had instead stuck to using grep as is their wont. Already like grizzled graybeard devs stuck in their ways. Anyway, I asked them to poke around and try it and let me know what they thought. So they took it for a test drive and had some feedback. Then I asked them to figure out when and why and how they could best use it, and then had one of them come up with the most token-dense, intuitive explanation (at least intuitive to their robot brain) to explain all that, for use in an AGENTS dot md file. Here’s what it came up with; feel free to steal it for your own AGENTS dot md (and btw, my verdict for now is that it’s worth keeping around for a while to see how much they end up using it, but I think it’s probably pretty useful-- if they can actually be convinced to use it, which is why these little blurbs are so critical, and why they should focus on convincing rather than ordering the agents! Moral suasion ain’t just for humans…): --- ### Morph Warp Grep vs Standard Grep Warp Grep = AI agent that greps, reads, follows connections, returns synthesized context with line numbers. Standard Grep = Fast regex match, you interpret results. Decision: Can you write the grep pattern? - Yes → Grep - No, you have a question → mcp__morph-mcp__warp_grep #### Warp Grep Queries (natural language, unknown location) "How does the moderation appeals flow work?" "Where are websocket connections managed?" "What happens when a user submits a post?" "Where is rate limiting implemented?" "How does the auth session get validated on API routes?" "What services touch the moderationDecisions table?" #### Standard Grep Queries (known pattern, specific target) pattern="fileAppeal" # known function name pattern="class.*Service" # structural pattern pattern="TODO|FIXME|HACK" # markers pattern="processenv" path="apps/web" # specific string pattern="import.*from [']@/lib/db" # import tracing #### What Warp Grep Does Internally One query → 15-30 operations: greps multiple patterns → reads relevant sections → follows imports/references → returns focused line ranges (e.g., l3-guardian.ts:269-440) not whole files. #### Anti-patterns | Don't Use Warp Grep For | Why | Use Instead | |------------------------|-----|-------------| | "Find function handleSubmit" | Known name | Grep pattern="handleSubmit" | | "Read the auth config" | Known file | Read file_path="lib/auth/..." | | "Check if X exists" | Boolean answer | Grep + check results | | Quick lookups mid-task | 5-10s latency | Grep is 100ms | #### When Warp Grep Wins - Tracing data flow across files (API → service → schema → types) - Understanding unfamiliar subsystems before modifying - Answering "how" questions that span 3+ files - Finding all touching points for a cross-cutting concern — You can see the whole short interaction in this gist: https://t.co/Z6jvFXDxub
Actually, I like the first one it came up with, too, I'm going to include both in my AGENTS dot md: --- ### Morph Warp Grep — AI-powered code search **Use `mcp__morph-mcp__warp_grep` for exploratory "how does X work?" questions.** An AI search agent automatically expands your query into multiple search patterns, greps the codebase, reads relevant files, and returns precise line ranges with full context—all in one call. **Use `ripgrep` (via Grep tool) for targeted searches.** When you know exactly what you're looking for—a specific function name, error message, or config key—ripgrep is faster and more direct. **Use `ast-grep` for structural code patterns.** When you need to match/rewrite AST nodes while ignoring comments/strings, or enforce codebase-wide rules. **When to use what** | Scenario | Tool | Why | |----------|------|-----| | "How is authentication implemented?" | `warp_grep` | Exploratory; don't know where to start | | "Where is the L3 Guardian appeals system?" | `warp_grep` | Need to understand architecture, find multiple related files | | "Find all uses of `useQuery(`" | `ripgrep` | Targeted literal search | | "Find files with `console.log`" | `ripgrep` | Simple pattern, known target | | "Rename `getUserById` → `fetchUser`" | `ast-grep` | Structural refactor, avoid comments/strings | | "Replace all `var` with `let`" | `ast-grep` | Codemod across codebase | **warp_grep strengths** * **Reduces context pollution**: Returns only relevant line ranges, not entire files. * **Intelligent expansion**: Turns "appeals system" into searches for `appeal`, `Appeals`, `guardian`, `L3`, etc. * **One-shot answers**: Finds the 3-5 most relevant files with precise locations vs. manual grep→read cycles. * **Natural language**: Works well with "how", "where", "what" questions. **warp_grep usage** ``` mcp__morph-mcp__warp_grep( repoPath: "/data/projects/communitai", query: "How is the L3 Guardian appeals system implemented?" ) ``` Returns structured results with file paths, line ranges, and extracted code snippets. **Rule of thumb** * **Don't know where to look** → `warp_grep` (let AI find it) * **Know the pattern** → `ripgrep` (fastest) * **Need AST precision** → `ast-grep` (safest for rewrites) **Anti-patterns** * ❌ Using `warp_grep` to find a specific function name you already know → use `ripgrep` * ❌ Using `ripgrep` to understand "how does X work" → wastes time with manual file reads * ❌ Using `ripgrep` for codemods → misses comments/strings, risks collateral edits
Want to learn how to build a lean, profitable, one-person business you love? I share a strategy you can implement immediately every Saturday morning in my newsletter. Always 4 minutes (or less) to read. Join us here: https://t.co/idNByMf6vn
The $10M Solopreneur | Helping 100,000+ burned-out corporate professionals build six-figure, one-person online businesses at https://t.co/27OAdtwR6D
6. 冷知识：巴菲特的父亲曾经是国会议员。因此本身就和政治走得很近，能知道很多政策内幕信息（因此可以安心长期拿着某些股票而不被美国政府算账）。 只是他的个人理念比较实诚和正能量，特别会讲安慰别人的话，尤其是面对恐惧和贪婪的人们讲话。 巴菲特的思维方式和背景，作为散户的你都学不会。
Research Scientist @Google | Previously PhD @nlp_usc, B.Eng @TsinghuaNLP 找工作、找面试题、改简历、模拟面试。关注： 创业（冷启动） ｜ 认知心理学｜智能体 ｜ 强化学习 building：https://t.co/A4YmEz9yqG
HARDCORE Hei is pretty obnoxious in the movie, so I can see it there's a pretty ruthless vibe throughout ironically this strengthens my recommendation great artists are rarely good men
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1
Russia has been degrading Ukrainian grid for like 3 years at this point. It's not a 4D strategy, it's very straightforward, and it's not suddenly going to work much better. I am irritated by people who've been watching it closer than me present the same grug stuff as novel info
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1