cablehead

kicking the tires on jev with 2048

You can watch these strategies play out here: https://jev-with-2048.ndyg.cross.stream

I finally got a chance to kick the tires on jev (jev-1.13.0). I thought it'd do really well playing 2048.

the strategy I used was: hand it the current board state and give it the option of up, down, left, right. I tried that 4 ways (the jev rows in the table).. each row links to the request I sent.

with just the board it does about as well as making random moves. it does best when code works out what each move would do to the board and jev picks from those.. then it's about as good as a fixed rule (e.g. always left if that moves anything, else down, else right, else up).

cablehead

2048: why two identical boards can have different scores

Two long games reached effectively the same position but finished with different scores: 1,811,320 vs 1,810,888, a gap of 432. The boards looked the same, so the scores "should" have matched. They don't, and there is no bug. Here is the full explanation, with the exact numbers.

The snapshot where the two boards coincide

There is a moment where both games show an identical set of 16 tiles (total

cablehead

nu-mcp: value reuse between agent turns

The usual MCP failure mode is a tool that returns a large value: the agent either floods its context window or has to re-run the command with | head to claw it back. Both options cost turns and lose information.

nu-mcp sidesteps this by storing every evaluation's full result in a ring buffer the agent can refer to by index on subsequent calls.

cablehead

python -c vs. nushell for JSON

Two ways to pull stargazers_count off the nushell repo's GitHub API response. Both verified against https://api.github.com/repos/nushell/nushell.

One field

bash + python -c:

cablehead

(
  interleave
    { generate {|_ = 0|
        let ev = (try { input listen --types [key] --timeout 200ms } catch { null })
        {out: (if $ev != null { {kind: key, event: $ev} }), next: 0}
      }
    }
    { generate {|_ = 0| sleep 1sec; {out: {kind: tick, event: (date now)}, next: 0} } }
)
| where $it != null

cablehead

with-dotenv {
    echo '{"role":"user","content":"Run this in the nu tool exactly: use greeter; greeter hi"}'
    | yoke --provider openrouter --model openai/gpt-oss-20b:free --tools nu -I ~/yoke-test/modules
  }

{"type":"tools","tools":[{"name":"nu","description":"Execute a Nushell script. Output is auto-converted to nuon -- do NOT add '| to nuon' or '| to json'. Pass structured data via 'input' (JSON) to avoid quoting issues -- it becomes $in in the pipeline.\n\nUse help liberally to learn how nushell works. Do NOT guess command names or flags -- discover them with help. If you encounter an error, you MUST use help related to your task before trying again.\n\nExamples:\n  {command: \"help where\"}\n  {command: \"help --find convert\"}\n  {command: \"$in | where price > 20\", input: [{name: \"Widget A\", price: 25.50}, {name: \"Gadget\", price: 15}]}\n  {command: \"seq 1 10 | each { |n| $n * $n }\"}"}]}
{"type":"agent_start"}
{"type":"turn_start"}
{"role":"user","content":[{"type":"text","text":"Run this in

cablehead

In the for_collect case, my understanding is the main goal of FilePath is to prevent clobbering files with save. If we've collected, it's now safe to overwrite our original file, so FilePath is cleared. The code has always been doing this, but the previous version tied that decision to content_type — it only cleared FilePath when content_type was also None. That coupling didn't make sense, so when I added the custom metadata field, I moved the collect-time cleanup into its own for_collect method on PipelineMetadata and made it always clear FilePath regardless of other fields.

cablehead

Connecting your app to Stellar for live CSS updates

Stellar is a local-first CSS generator with an interactive web editor. stellar serve serves your generated CSS and exposes a /live-refresh SSE endpoint that pushes stylesheet updates via Datastar.

stellar serve  # http://localhost:7331

cablehead

xs Naming Conventions RFC

Proposing consistent naming for stream position concepts in xs.

The Problem

Current naming mixes metaphors:

• .head returns the newest frame (like Git HEAD)
• --tail skips to the end (like Unix tail)
• But resume_from: "head" means start from the beginning

cablehead

Exploring schema validation for nushell/http-nu

We want to validate records against a schema (like JSON Schema). Nushell already has type syntax for command signatures:

def foo []: record<
  name: string,
  age: int,
  email?: string,
  tags: list<string>,