Skip to content

Instantly share code, notes, and snippets.

View HanClinto's full-sized avatar

Clint Herron HanClinto

View GitHub Profile
@HanClinto
HanClinto / qwen3.8flashnext_response.md
Created October 22, 2026 03:01
Asking Qwen 3.8 Flash-Next to review my "CollectorVision" repository

Prompt

Are you able to browse the web? I’m curious what you think of this repository: https://github.com/HanClinto/CollectorVision/

Response

Yes — I can fetch public web pages and run searches. Let me take a look at that repo.

@HanClinto
HanClinto / 00-START-HERE.md
Created September 5, 2026 13:49
WebMCP Challenge: 2,500-entry review, revised top ten, detailed scores and sorting guide (2026-09-05)

WebMCP Challenge Evaluation

Independent AI-assisted evaluation prepared September 5, 2026, published at HanClinto's request. Not official judging results or calibrated win probabilities.

Start Here

  • Full report: revised top ten, before/after comparison, alternatives and limitations.
  • Google Sheets: all project feedback: the spreadsheet copy supplied by the owner. This publication does not update or verify that copy's contents or permissions.
  • Detailed top-ten CSV: final scores, four final-score explanations, strengths, weaknesses, evidence level and comparison rationale. Use the Raw/download control for import into Sheets.
  • Finalist explanations: the same detailed analysis in readable Markdown.

WebMCP Challenge: Predicted Top 10

Assessment date: 2026-09-05. This is an independent, provisional prediction, not an official judging result.

Recommendation

Jig is my strongest single pick; JupyterLite WebMCP and Roque Nights are the next two. The ten recommended projects below are my best-supported selection from the public evidence reviewed. The contest has ten equal winning slots, not a single grand-prize winner, so the useful prediction is membership in that group, not the exact internal ordering.

I collected all 2,500 unique projects returned by the supplied search, across 105 pages, and saved all 2,500 full descriptions. Every description received automated evidence screening. I read selected summaries and fuller descriptions, collected public source snapshots for a 46-project shortlist, and opened all 46 live URLs. All returned HTTP 200, but a password gate or landing page is not proof of a working product. Ten projects have explicit four-criterion scorecards. **This is not a

@HanClinto
HanClinto / README.md
Last active August 11, 2026 05:35
llama.cpp grammar closure stack-copy benchmark

llama.cpp grammar closure stack-copy benchmark

This benchmark compares two clean llama.cpp worktrees at the same base revision:

  • baseline: unmodified 030ebb558a5820b444a8f836ed5cdd46c9b4bd7a;
  • optimized: 2703482ef97cf0aa93e48b0f3e7c8c2c7490b0c0, based on the baseline with one change in llama_grammar_advance_stack().

The optimized code combines closure-stack membership testing and insertion, moves the pending vector into the existing seen set, and processes the canonical set-owned stack by reference. It does not change grammar state or accepted output stacks.

Run

Description

Tool to calculate the EV of an intentional draw for a player in a Swiss-pairings tournament (such as Magic: The Gathering)

NOTE: WIP -- it is not complete, and this is just a basic structure for keeping track of the tournament structure, as well as the calculations for a players' own points, as well as the various tiers of tiebreakers.

The intention is to build this basic framework into a tool that can do Monte Carlo style simulations (either exhaustive or sampled) to calculate the EV of intentional draws.

Currently a bit stuck on whether or not I should track results at the Game level, or simply at the Match level. Either way, I'm a little stuck on knowing how to determine if a Match is complete via intentional draw, and it's something that I might just have to special-case. Regardless, it's a tricky situation because I'm also trying to write this to support multiplayer, as well as mixed draw+loss resulst being in a single 4-player game (I.E., two players lose, and the final two players tak

@HanClinto
HanClinto / open_tunnel.sh
Last active May 15, 2023 21:06
Open a VSCode tunnel on an arm64 device
# If `code` does not exist in the local directory, then download it.
# https://code.visualstudio.com/sha/download?build=stable&os=cli-alpine-arm64
if [ ! -f code ]; then
echo "Downloading code..."
wget -q "https://code.visualstudio.com/sha/download?build=stable&os=cli-alpine-arm64" -O code.tar.gz
tar -xzf code.tar.gz
rm code.tar.gz
fi
@HanClinto
HanClinto / Response from GPT-J
Created August 11, 2021 17:58
Prompt to GPT-J 6b
A:
I can't tell you where the problem lies, but you should:
make sure you have S3-specific permissions in place for the role you are using for your S3 work, that is, (a) access and (b) full control
make sure you have S3-specific policies attached to the S3 resources in question, that is, (a) access and (b) full control
make sure you are using the correct role
make sure that the resource exists
make sure you have the right policy
import aicrowd_gym
import gym
import nle
import numpy as np
import collections
from nle import nethack as nh
def getPrintableChar(a):
non_printable_chars = ['NULL', 'START_OF_HEADING', 'START_OF_TEXT', 'END_OF_TEXT', 'END_OF_TRANSMISSION', 'ENQUIRY', 'ACKNOWLEDGE', 'BELL', 'BACKSPACE', 'HORIZONTAL_TAB', 'LINE_FEED', 'VERTICAL_TAB', 'FORM_FEED', 'CARRIAGE_RETURN', 'SHIFT_OUT', 'SHIFT_IN', 'DATA_LINK_ESCAPE', 'DEVICE_CONTROL_1', 'DEVICE_CONTROL_2', 'DEVICE_CONTROL_3', 'DEVICE_CONTROL_4', 'NEGATIVE_ACKNOWLEDGE', 'SYNCHRONOUS_IDLE', 'END_OF_TRANSMISSION_BLOCK', 'CANCEL', 'END_OF_MEDIUM', 'SUBSTITUTE', 'ESCAPE', 'FILE_SEPARATOR', 'GROUP_SEPARATOR', 'RECORD_SEPARATOR', 'UNIT_SEPARATOR']
if a < 32:
## Actions: `env._actions`
### Actions sorted by action index:
| Index | Action | ASCII | Character |
| ----- | ------ | ----- | --------- |
| 0 | CompassDirection.N | '`k`' | 107 |
| 1 | CompassDirection.E | '`l`' | 108 |
| 2 | CompassDirection.S | '`j`' | 106 |
| 3 | CompassDirection.W | '`h`' | 104 |
@HanClinto
HanClinto / highresslowmo.ipynb
Created August 6, 2020 18:49
HighResSlowMo.ipynb
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.