Replace README and add combined banner

.gitattributes

@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 modHacker-F16.gguf filter=lfs diff=lfs merge=lfs -text
+assets/modai-modhacker-banner.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -5,7 +5,6 @@ language:
 pipeline_tag: text-generation
 tags:
   - gguf
-  - qwen3.5
   - uncensored
   - lower-refusal
   - coding
@@ -17,39 +16,37 @@ tags:
 ---
 
 <p align="center">
-  <img src="assets/modhacker-banner.svg" alt="modHacker by STUDIOBRN / We Are The Art Makers" width="100%">
+  <img src="assets/modai-modhacker-banner.png" alt="modAI and modHacker" width="100%">
 </p>
 
-<p align="center">
-  <strong>LOCAL AI · LOWER REFUSAL · CODE · SYSTEMS</strong><br>
-  <a href="https://WeAreTheArtMakers.com">We Are The Art Makers</a> · STUDIOBRN
-</p>
+<p align="center"><strong>LOCAL AI · CODE · SYSTEMS · REASONING</strong><br>
+STUDIOBRN / <a href="https://WeAreTheArtMakers.com">We Are The Art Makers</a></p>
 
 # modHacker
 
-**modHacker** is an experimental **uncensored / lower-refusal** local model distribution for coding, debugging, system analysis, authorized cybersecurity research, reasoning, terminal workflows and agent development. The release packages Qwen3.5-family F16 GGUF weights with an optional STUDIOBRN / We Are The Art Makers Ollama system configuration.
+**modHacker** is a locally runnable **uncensored, lower-refusal, abliterated-style** AI model release for developers who want to explore coding, debugging, system analysis, terminal workflows and agentic development on their own hardware. It ships as a single F16 GGUF file, with an optional Ollama configuration created by STUDIOBRN / We Are The Art Makers.
 
-**About “abliterated”:** modHacker belongs to the broader *abliterated / uncensored* model space because an upstream component was designed for refusal removal. The published lineage describes a weight merge; there is no documented separate abliteration pass on modHacker's GGUF weights. “Uncensored” describes the intended lower-refusal behavior, not a measured refusal rate for this release. See [Provenance](#provenance).
+> “Abliterated-style” describes the release's intended lower-refusal behavior. This distribution does not claim an independently verified refusal rate or a new abliteration procedure performed by STUDIOBRN.
 
-| Release detail | Value |
+| Detail | Release |
 | --- | --- |
-| Model family | Qwen3.5, 4B class (approximately 4.3B reported by Ollama) |
-| Distribution | `modHacker-F16.gguf`, one F16 GGUF file, approximately 8.67 GB |
-| Ollama context setting | `num_ctx 16384` in the included `Modelfile` |
-| Focus | Local coding, debugging, system analysis, reasoning, authorized security research |
-| Behavioral aim | Lower refusal / uncensored; independently measured refusal rate unavailable |
-| Maintainer and packaging | STUDIOBRN / We Are The Art Makers |
-| License | Apache License 2.0 for the underlying model; see [LICENSE](LICENSE) and [NOTICE](NOTICE) |
+| Format | F16 GGUF (`modHacker-F16.gguf`) |
+| Size | Approximately 8.67 GB |
+| Model class | Approximately 4B parameters |
+| Local runtime | Ollama; other compatible GGUF runtimes |
+| Included Ollama setting | 16K context (`num_ctx 16384`) |
+| Focus | Coding, debugging, reasoning, system analysis and local AI agents |
+| Distribution and configuration | STUDIOBRN / We Are The Art Makers |
 
-## Get started
+## Run locally
 
-### Run the published GGUF directly with Ollama
+Run the GGUF directly with Ollama:
 
 ```bash
 ollama run hf.co/studiobrn/modHacker:F16
 ```
 
-This uses the published GGUF. To add the STUDIOBRN system prompt and 16K context configuration, download the file and build the named local model:
+To use the included **modHacker identity, system prompt and 16K context** configuration:
 
 ```bash
 hf download studiobrn/modHacker modHacker-F16.gguf Modelfile --local-dir ./modHacker
@@ -58,28 +55,16 @@ ollama create modHacker -f Modelfile
 ollama run modHacker
 ```
 
-The second route requires enough free storage for the download and Ollama import. `Modelfile` points to `./modHacker-F16.gguf`; run `ollama create` from the directory that contains both files. The repository's GGUF and Modelfile are separate files: the system prompt and license notice are not baked into the weights.
-
-### Other GGUF runtimes
-
-Compatible recent versions of llama.cpp and other GGUF runtimes can load the file. Tool calling and image support depend on the runtime, its chat template, and any required multimodal assets; this release has no separately listed `mmproj` file. The user-facing focus of this package is text and local agent workflows.
-
-## Configuration
-
-The included [`Modelfile`](Modelfile) sets `num_ctx 16384` and includes the requested modHacker identity and operating principles. It asks the assistant to be precise, practical and candid about tool use and uncertainty, and to support legitimate software and defensive security tasks. A system prompt guides behavior; it does not train or alter the GGUF weights. Memory use rises with context length.
-
-## Provenance
-
-The weights originate from the [AESIR-Hacker-Micro F16 GGUF conversion](https://huggingface.co/mradermacher/AESIR-Hacker-Micro-GGUF). The [AESIR model card](https://huggingface.co/aesir-unlimited/AESIR-Hacker-Micro) describes a 40/60 merge of `Qwen/Qwen3.5-4B` and `rodrigomt/Qwen3.5-4B-Uncensored-Aggressive`, the latter converted from HauhauCS's aggressive uncensored release. The AESIR card reports no independent benchmark for the resulting merge. STUDIOBRN's work here is configuration, packaging and branding; this release does not claim a new fine-tune, new training data, or a newly performed abliteration.
+Run `ollama create` from the directory containing both the GGUF and `Modelfile`. The direct Hugging Face run loads the GGUF; the `ollama create` route applies this repository's custom system prompt.
 
-The `base_model` metadata field is intentionally left empty. Provenance and upstream attribution are documented here and in [NOTICE](NOTICE).
+## Built for practical local workflows
 
-## Intended use and limits
+The included [`Modelfile`](Modelfile) defines modHacker's working style: concise technical answers, clear assumptions, maintainable code, careful debugging and honest reporting of tool use. The model can be connected to an agent harness for tool workflows; actual tool execution depends on that harness and runtime.
 
-Use it for locally run coding assistance, testing ideas, maintenance, reverse engineering within authorization, and defensive research. Check generated commands and code before executing them, especially when they affect live systems. No independent benchmark, refusal-rate test or security evaluation of this specific GGUF package is published here. Text generated by the model can still be wrong.
+The companion [**modAI CLI**](https://github.com/WeAreTheArtMakers/modAI) explores multiple agent roles, local model use and coding workflows. The visual above brings the model and CLI together under one local AI project.
 
-## License and credits
+## License
 
-The underlying weights are distributed under the **Apache License 2.0**. Keep upstream notices and attribution intact when redistributing. The `modHacker` name, system prompt, Ollama configuration, packaging and branding are credited to **© 2026 We Are The Art Makers / STUDIOBRN**. The [NOTICE](NOTICE) records this without replacing any upstream license or claim of ownership over upstream weights.
+The underlying model is distributed under **Apache License 2.0**. See [LICENSE](LICENSE) and [NOTICE](NOTICE) for the license text, distribution credits and upstream attribution. The modHacker name, configuration, system prompt and packaging are credited to **© 2026 We Are The Art Makers / STUDIOBRN**.
 
-**Project:** [WeAreTheArtMakers.com](https://WeAreTheArtMakers.com)
+[WeAreTheArtMakers.com](https://WeAreTheArtMakers.com) · [modAI on GitHub](https://github.com/WeAreTheArtMakers/modAI)