AI & ML interests
Post-training, model behavior and transparent evaluation. Research from alphaBell.
Recent Activity
View all activity
Articles
Add the community-inspired deterministic replay lab
#7 opened about 1 hour ago
by 
alphabell-research
Feature the new MOSS and DeepSeek tools
#3 opened about 2 hours ago
by alphabell-research
alphabell-research
published a
Space about 2 hours ago
Publish the source-checked prompt case explorer
#1 opened about 2 hours ago
by alphabell-research
alphabell-research
published an article about 2 hours ago
Article
Draw MOSS-Image layouts without doing coordinate arithmetic
alphabell
• alphabell-research
published a
Space about 3 hours ago
Launch the reference-checked MOSS layout editor
#1 opened about 3 hours ago
by alphabell-research
Add live app links, screenshots and social previews
#4 opened about 3 hours ago
by alphabell-research
Add live app links, screenshots and social previews
#6 opened about 3 hours ago
by alphabell-research
Link the public request-feedback thread from the app
1
#3 opened about 5 hours ago
by alphabell-research
Link the public repair challenge from the app
1
#5 opened about 5 hours ago
by alphabell-research
Suggest a request check or report a version difference
#2 opened about 5 hours ago
by alphabell-research
Repair challenge: the smallest patch that brings a game back
#4 opened about 5 hours ago
by alphabell-research
Feature Request Doctor alongside the Flint repair projects
1
#2 opened about 6 hours ago
by alphabell-research
alphabell-research
posted an update about 6 hours ago
Post
39
In Rec MLX v0.1.0, thinking_token_budget can look valid and have no effect: the server removes it. And max_tokens: 15000.0 is a Python float where an integer is required.
We built Flint Request Doctor to inspect these requests locally:
alphabell/Flint-Request-Doctor
It distinguishes ignored fields from rejected settings, preserves original number spellings and large integer values when copying, and links diagnostics to pinned source. Its decisions match upstream Python request validation in 149 comparison cases.
What other Rec or Flint integration traps should we add?
We built Flint Request Doctor to inspect these requests locally:
alphabell/Flint-Request-Doctor
It distinguishes ignored fields from rejected settings, preserves original number spellings and large integer values when copying, and links diagnostics to pinned source. Its decisions match upstream Python request validation in 149 comparison cases.
What other Rec or Flint integration traps should we add?
- 1 reply
alphabell-research
updated a
collection about 6 hours ago
alphabell-research
published a
Space about 6 hours ago
Build a source-pinned request checker for Rec Flint MLX
2
#1 opened about 6 hours ago
by alphabell-research
Keep phone controls visible and publish reproducible browser checks
1
#3 opened about 7 hours ago
by alphabell-research
alphabell-research
updated a
collection about 7 hours ago