Restore thumbnail and update Jack 3.8 model card

README.md

@@ -1,4 +1,4 @@
----
+---
 tags:
 - 27b
 - gguf
@@ -25,6 +25,10 @@ pipeline_tag: text-generation
 ---
 
 # Jack-3.8-27B-Coder-16GB-VRAM
+
+<p align="center">
+  <img src="https://huggingface.co/JackAgentLead/Jack-3.8-27B-Coder-16GB-VRAM/resolve/main/jack-3.8-coder.png" alt="Jack 3.8 Coder" width="600">
+</p>
 
 > **27B coding intelligence. Structured cognitive control. One 16GB GPU.**
 
@@ -158,7 +162,7 @@ Conceptually:
 
 ```text
 LONG CONTEXT / NATIVE REASONING
-────────────────────────────────────────────
+────────────────────────────────────────────
 old state
 new state
 hypotheses
@@ -170,17 +174,17 @@ corrections
 more reasoning
 more reasoning
 more reasoning
-                         │
-                         ▼
+                         │
+                         â–¼
 JACK XML REBINDING
-────────────────────────────────────────────
+────────────────────────────────────────────
 <workspace_state>   operative state
 <grounded_source>   evidence that controls the conclusion
 <anchor_fact>       exact critical variables and identities
 <deterministic_check> verified external results when applicable
 <pitfall_check>     active falsification / failure boundaries
-                         │
-                         ▼
+                         │
+                         â–¼
 FINAL OUTPUT / ACTION
 ```
 
@@ -247,9 +251,9 @@ This targets a common model failure:
 
 ```text
 plausible prior knowledge
-        ↓
+        ↓
 silently replaces
-        ↓
+        ↓
 the evidence actually supplied in context
 ```
 
@@ -372,36 +376,36 @@ A typical workflow can look like:
 PLAN / NOVEL PROBLEM
 Native thinking: ON
 Jack XML: ACTIVE
-        │
-        ▼
+        │
+        â–¼
 expansive planning, search, falsification
 
 ROUTINE EXECUTION
 Native thinking: OFF
 Jack XML: ACTIVE
-        │
-        ▼
+        │
+        â–¼
 aligned execution under anchored state
 
 NEW FAILURE / UNCERTAINTY
 Native thinking: ON
 Jack XML: ACTIVE
-        │
-        ▼
+        │
+        â–¼
 deep debugging and hypothesis search
 
 CONTINUED EXECUTION
 Native thinking: OFF
 Jack XML: ACTIVE
-        │
-        ▼
+        │
+        â–¼
 efficient implementation
 
 INDEPENDENT REVIEW
 Native thinking: ON
 Jack XML: ACTIVE
-        │
-        ▼
+        │
+        â–¼
 adversarial audit and recalibration
 ```
 
@@ -439,19 +443,19 @@ Observed trajectory:
 
 ```text
 native thinking: OFF
-        ↓
+        ↓
 <workspace_state>
-        ↓
+        ↓
 <grounded_source>
-        ↓
+        ↓
 initial <anchor_fact> values
-        ↓
+        ↓
 <pitfall_check>
-        ↓
+        ↓
 detect inconsistency
-        ↓
+        ↓
 supersede incorrect anchors
-        ↓
+        ↓
 correct final authoritative state
 ```
 
@@ -509,16 +513,16 @@ It is preserving the correct engineering state across:
 
 ```text
 requirement
-→ plan
-→ implementation
-→ failure
-→ diagnosis
-→ revised plan
-→ new artifact
-→ test
-→ external review
-→ correction
-→ final verification
+→ plan
+→ implementation
+→ failure
+→ diagnosis
+→ revised plan
+→ new artifact
+→ test
+→ external review
+→ correction
+→ final verification
 ```
 
 Every earlier state may remain inside the context.
@@ -545,13 +549,13 @@ The review target is broader than ordinary code inspection:
 
 ```text
 original requirement
-        ↕
+        ↕
 derived plan
-        ↕
+        ↕
 implementation
-        ↕
+        ↕
 tests
-        ↕
+        ↕
 claimed verification
 ```
 
@@ -572,7 +576,7 @@ They demonstrate the kind of auditing behavior Jack is designed to perform.
 
 ---
 
-# State-Preserving Plan → Execute → Review
+# State-Preserving Plan → Execute → Review
 
 Jack 3.8 has also been tested in workflows where native reasoning mode changes between stages.
 
@@ -581,17 +585,17 @@ A representative pattern:
 ```text
 PLANNING
 Native thinking: ON
-        ↓
+        ↓
 Jack reconstructs and anchors the authoritative plan
 
 EXECUTION
 Native thinking: OFF
-        ↓
+        ↓
 Jack carries out the established plan under active XML alignment
 
 REVIEW
 Native thinking: ON
-        ↓
+        ↓
 Jack independently falsifies the result
 ```
 
@@ -627,13 +631,13 @@ The intended relationship is:
 
 ```text
 model hypothesis
-        ↓
+        ↓
 deterministic check
-        ↓
+        ↓
 verified result
-        ↓
+        ↓
 Jack rebinds verified result into operative state
-        ↓
+        ↓
 later reasoning must use, explain, or challenge the evidence explicitly
 ```
 
@@ -807,7 +811,7 @@ The evaluations examined:
 
 - planning precision
 - preservation of original requirements
-- plan → execution fidelity
+- plan → execution fidelity
 - malformed-input handling
 - test quality
 - independent review
@@ -1030,7 +1034,7 @@ Jack is an inference-time cognitive control architecture integrated with a capab
 | Field | Value |
 |---|---|
 | Model | `Jack-3.8-27B-Coder-16GB-VRAM` |
-| Author | Jonathan Michael Langford — Independent Researcher, Lead Architect, The Jack Project |
+| Author | Jonathan Michael Langford — Independent Researcher, Lead Architect, The Jack Project |
 | Contact | `mlangford75@protonmail.com` |
 | Project / Research | https://github.com/mlangford75-lgtm/mlangford75-lgtm |
 | GGUF | `Jack-3.8-27B-Coder-16GB-VRAM.gguf` |
@@ -1085,3 +1089,4 @@ Jack asks a practical question:
 > **How much effective agentic intelligence can be recovered from local hardware by organizing cognition better rather than relying only on more parameters, more context, or permanently enabled deliberation?**
 
 That is the research program.
+