Re: [NonGNU] new: hermes

On 2026-08-09 22:37, Richard Stallman wrote:
> Jean Louis wrote:
> 
>   > Let's not forget that Hermes is used only by small percentage of 
> users
>   > running fully free LLMs on their computers,
> 
> I am having trouble understanding the meaning of that statement.
> It seems to say that one quantity A is a small fraction of another 
> quantity B.
> 
> Could you please state carefully the meaning of quantity A, and the 
> meaning
> of quantity B?

Quantity A refers to the small number of Hermes Agent users who run 
fully free, locally-hosted LLMs — a group that remains small because 
setting up local inference (models, weights, hardware, dependencies) is 
far more burdensome than signing up for a commercial SaaS LLM API.

Quantity B refers to the large majority of Hermes Agent users who 
instead use remote, non-free SaaSS LLMs — the path the installer 
defaults to, promotes, and makes easiest.

Emacs package supporting Hermes software agent, doesn't really IMHO push 
software freedom, but opens up good door for Emacs users to simply get 
hooked up on remote SaaSS LLMs, which already control users by large.

Emacs package interacts with the Hermes agent.

Hermes agent by default funnels users toward non-free SaaS LLMs 
companies (as it was made for that purpose).

Install process of Hermes agent asks users to sign up on the portal 
https://portal.nousresearch.com/manage-subscription to get subscription 
to SaaSS LLMs.

Yet, Eli said it is easy to avoid that by simply pressing Ctrl-C -- yet 
it is nowhere written or warned or given to user that user could press 
Ctrl-C to avoid that.

And I remember time when Hermes was advertised without any "portal" -- 
so there was no enforced subscription (even if one of four subscriptions 
is for zero money, user is driven to portal to use remote API).

Then during the process of installation the default option is SaaSS LLM 
from Nous company, and not the local model.

Frankly, commercial and proprietary LLMs are driving the whole industry 
of such locally installed software agents. While software is made free, 
under free license, it is by large being used to access SaaSS LLMs.

Most Developers Call AI Data With APIs and A2A - The New Stack:
https://thenewstack.io/most-developers-call-ai-data-with-apis-and-a2a/

AI Agents Are Your API's Biggest Consumer. Do They Care About Good 
Design? - DEV Community:
https://dev.to/aws/ai-agents-are-your-apis-biggest-consumer-do-they-care-about-good-design-31l2

LLM (Large Language Model) Statistics 2026 – Users, Adoption & Revenue:
https://www.getpanto.ai/blog/llm-statistics

ChatGPT	900 million weekly active users
Google Gemini	900 million monthly active users
Google AI Overviews	2.5 billion monthly users
Google AI Mode	1 billion+ monthly users
Claude	300,000+ business customers

While those statistics do not show how many people run locally some LLM, 
we can see that industry is based on sales of SaaSS LLMs and most models 
considered "best" are also proprietary in the same time, which means 
users cannot duplicate it on their own computers.

Hermes agent is software that by large it is designed to use non-free 
SaaSS LLMs from proprietary companies.

It has been designed to make money, which I don't object, but I see such 
a resistance on mailing lit to talk the truth about it.

I don't mind that Emacs package is included or not, as I don't have 
authority on that. But as long as Eli doesn't reject my email with 
excuse how it is not appropriate, I will be stating my opinion as I am 
free software user, and not that pragmatic one.

On this mailing list few people justify including support in Emacs 
packages to such software like Hermes, for reason that 1 (one) option 
allows it running the local LLM.

Yet other options are overwhelming supporting SaaS LLMs as seen below 
from the setup screen:

Select provider:
   ↑↓ navigate  ENTER/SPACE select  ESC cancel

  → (●) Nous Portal (Everything your agent needs, 300+ models with 
bundled tool use)
    (○) Fireworks AI (OpenAI-compatible direct model API)
    (○) OpenRouter (Pay-per-use API aggregator)
    (○) Mixture of Agents (named presets; aggregator acts after reference 
models)
    (○) NovitaAI (Cloud: Model API, Agent Sandbox, GPU Cloud)
    (○) LM Studio (Local desktop app with built-in model server)
    (○) Anthropic (Claude models via API key or Claude Code)
    (○) OpenAI ▸ (Codex CLI or direct OpenAI API)
    (○) Qwen ▸ (Qwen Cloud / DashScope, Coding Plan & Qwen CLI OAuth)
    (○) xAI Grok ▸ (Direct API or SuperGrok / Premium+ OAuth)
    (○) Xiaomi MiMo (MiMo-V2.5 and V2 models: pro, omni, flash)
    (○) Tencent TokenHub (Hy3 Preview via tokenhub.tencentmaas.com)
    (○) NVIDIA NIM (Nemotron models via build.nvidia.com or local NIM)
    (○) GitHub Copilot ▸ (GitHub token API or copilot --acp process)
    (○) Hugging Face Inference Providers
    (○) Google AI Studio (Native Gemini API)
    (○) Google Vertex AI (Gemini via GCP; OAuth2 service account or ADC, 
GCP billing/quotas)
    (○) DeepSeek (V3, R1, coder, direct API)
    (○) Z.AI / GLM (Zhipu direct API)
    (○) Kimi / Moonshot ▸ (Coding Plan, Moonshot global & China 
endpoints)
    (○) StepFun Step Plan (Agent / coding models via Step Plan API)
    (○) MiniMax ▸ (Global, OAuth Coding Plan & China endpoints)
    (○) Ollama Cloud (Cloud-hosted open models, ollama.com)
    (○) Arcee AI (Trinity models, direct API)
    (○) GMI Cloud (Multi-model direct API)
    (○) Kilo Code (Kilo Gateway API)
    (○) OpenCode ▸ (Zen pay-as-you-go or Go subscription)
    (○) AWS Bedrock (Claude, Nova, Llama, DeepSeek; IAM or API key)
    (○) Azure Foundry (OpenAI-style or Anthropic-style endpoint, your 
Azure AI deployment)
    (○) Vercel AI Gateway (Multi-model aggregator)
    (○) Actual Computer - hosted inference via api.actual.inc, or local 
offline inference via ACTUAL_BASE_URL
    (○) custom (direct API)
    (○) DeepInfra — 100+ open models, pay-per-use
    (○) Upstage (Solar API)
    (○) Custom endpoint (enter URL manually)
    (○) Configure auxiliary models...
    (○) Leave unchanged

Yet none of participants on the mailing lists so far confirmed to 
actually run the local LLMs.

It seems that even when evidence shows the Emacs package for Hermes 
Agent will, by its design and defaults, push the vast majority of users 
toward SaaSS LLMs, the mere fact that it can support a local LLM is 
considered sufficient for its inclusion or promotion. However, even 
within NonGNU, I do not believe this genuinely serves the user's 
freedom. If the package were to include a prominent warning and guide 
users to understand the nature of SaaSS—perhaps by pointing them to "Who 
Does That Server Really Serve?"—then I would consider it a beneficial 
addition that respects and informs user autonomy.

I find it far more beneficial for free software promotion and for the 
GNU project to host and promote a fully free software agents (also 
tagged as "Local-First") and Emacs package that does not, by its very 
architecture and defaults, funnel users toward non-free SaaSS LLM 
companies. The goal should be to empower users to run their own 
computing with their own free software, on their own machines, using the 
free hardware and software stacks that are now becoming available, such 
as the Intel B60 GPU with SYCL/Vulkan (low-level compute APIs) and 
llama.cpp (LLM inference software using those APIs) running fully free 
models like Apertus (Apertus | Swiss AI:
https://www.swiss-ai.org/apertus).

-- 
Jean Louis