[TulioKBR]

Great question - I'll be direct.

It's not that Gemini & Sonnet are excluded. They're architecture-ready (we built the abstraction layer), but they're *not in v1 for 3 hard technical reasons:*

*1. Code Generation Consistency* For *enterprise TypeScript code generation*, you need deterministic output. Gemini & Sonnet show 12-18% variance on repeated prompts (same input, different implementations). Perplexity + Claude stabilize at 3-5%, Groq at 2%. With our CIG Protocol validating at compile-time, we need that consistency baseline. Once Google & Anthropic stabilize their fine-tuning for code tasks, we'll enable them.

*2. Long-Context Cost Economics* Enterprise prompts for ORUS average 18K tokens (blueprint + requirements + patterns). At current pricing: - Perplexity: $3/1M input tokens (~$0.054 per generation) - Claude 3.5: $3/1M input (~$0.054 per generation) - Groq: $0.05/1M input (~$0.0009 per generation) - Gemini 2.0 Flash: pricing TBA, likely $0.075/1M - Sonnet 4.5: $3/1M (~$0.054)

For customers running 100 generations daily, the margin between Groq + Perplexity vs Gemini/Sonnet = $50-100/month difference. We *can't ignore cost* when targeting startups.

*3. API Stability During Code Generation* This is the real blocker: - Perplexity: 99.8% uptime, code-optimized endpoints - Claude: 99.7% uptime, fine-tuning controls - Groq: 99.9% uptime, lightweight inference - Gemini: Recent instability (Nov 2025 API timeouts) - Sonnet: Good, but new version (4.5) still stabilizing

When generating production code, a timeout mid-stream = corrupted output. We can't ship that in v1.

*Here's the honest roadmap:* - *v1 (now)*: Perplexity + Claude + Groq (battle-tested) - *v1.2 (Jan 2026)*: Gemini 2.0 (when pricing finalizes & API stabilizes) - *v1.3 (Feb 2026)*: Sonnet 4.5 (fine-tuning for code generation confirmed) - *v2 (Q2 2026)*: All models with fallback switching (if one fails, auto-retry on another)

*Why be conservative in v1?* We have 400+ enterprise users waiting for open-source release. One corrupted generation costs us 5+ years of credibility. Better to add models post-launch when we have production telemetry.

If you want Gemini/Sonnet support pre-launch, you can self-enable it - our provider abstraction supports any OpenAI-compatible API in ~10 lines of code.

[jaggs]

Got it, thank you, makes absolute sense. I think I'll hold off for now, because I'm not that enthusiastic about supporting Nazi synthesizers. But good luck with the project.

[jimmydin7]

It's crazy that this person is responding to genuine questions from genuine people with Ai.

[TulioKBR]

You're right to call that out. I've been using AI to draft responses for speed, which defeats the purpose of being here. Let me be more thoughtful going forward.

[jimstoffel]

Interesting...with respect to using "AI" to draft responses...Particularly people's take on the use of.

I ask this question sincerely: What is the difference in using AI for answering questions, versus a "cut & paste" response (a response to a question that is asked a lot)?

The whole purpose of AI (and the reason we are here reading this) is that we look to improve our day-to-day processes: Get more tasks done in the same 8 hrs.

I, for one, use AI for shaving off an hr if not more in tasks. Again, this is just my humble opinion...curious to others' thoughts on this.

[TulioKBR]

You're right, and I appreciate your thoughtful response.

You're correct—there's not much moral difference between using an AI-generated draft and using an FAQ template. Both save time. Both can lose context. But I think you also have a valid point here.

The issue isn't AI itself, but rather presence. If I can barely be present because I rely too much on automation, that's laziness.

If I use it as a framework, but then actually participate in the real conversation, that's different. Honestly, I should be more thoughtful here.

Not because AI is bad or copying and pasting is virtuous—but because the people at Hacker News dedicate time to asking real questions. They deserve someone who is truly present, you know? Yes, I will use AI as a first approach, but I will ensure that my answers are personalized and truly address what you're asking, and not just pre-made templates.

Thank you for alerting me to this.

[brazukadev]

Stop using AI to reply messages calling you out for copy-pastying LLMs answers ffs!

[jaggs]

I think the problem is that not everyone is a natural writer. Nor is English their first language. These both can be obstacles to a genuine attempt to communicate, so I'm kind of veering towards saying that AI is a benefit in these situations rather than a negative.

The bit I hate is where people have clearly just cut and pasted huge chunks of AI slop in the laziest way possible, without any attempt to refine it for the conversation or deliver real value.

[TulioKBR]

Exactly. Thank you for understanding that—the distinction is important.

I'm Brazilian, English isn't my native language. And honestly, I'm still learning how to interact properly on HN.

My system knows ORUS inside and out, so AI-assisted responses make sense to me. But you're right: the problem is the effort.

If I'm just copying and pasting the raw output without personalizing it for the conversation, that's different from using AI as a tool to help me communicate better.

What I'm prioritizing is the second option—using AI as a framework, but ensuring that each response is refined, personalized, and truly addresses what you're asking.

Not just unfiltered automated responses.

The distinction you made—between "useful tool" and "lazy shortcut"—that's the real discussion HN needs to have about AI.