Hamish from A-Tech

Anyone want Blackwell GPUs?

We're stocking them in a few days

Comment below

1 week ago | [YT] | 0

Hamish from A-Tech

Project Infra is finally here.


It began with a simple idea – giving everyone the power to own their own AI.


Countless hours of brainstorming turned into late‑night sketches of open‑source tools, and relentless work forged partnerships with Australia’s leading IT firms and computer suppliers.


Together, we built Project Infra.


A movement that brings the best‑in‑class open source AI capabilities to the masses—zero ongoing costs, unlimited use.


Our next milestone: the Open Source AI Builder’s Club, a global community of creators matched with businesses that need their talent.


Then came Cadent AU, an AI‑governance partner, ensuring our enterprise‑level consulting meets the highest standards.


All of this culminated in a moment we’ve all been waiting for:


We have officially launched the Project Infra store! 🎉


Step into the shop and discover ready‑to‑ship computers, each fine‑tuned by our specialist team into true AI powerhouses.


Visit us at: shop.atechds.com/

Limited stock for now. Working hard to add more every day!


Welcome to the next chapter of AI ownership.

4 months ago | [YT] | 0

Hamish from A-Tech

Going live now!!!

4 months ago | [YT] | 0

Hamish from A-Tech

🚀 10 Bedroom‑Based Business Ideas You Can Launch TODAY with the new gpt‑oss:120b 🚀

(All you need is a modest 16 GB‑RAM PC (or a single 80 GB GPU) and Ollama – the model runs locally, so you keep data private and costs near‑zero.)





1️⃣ AI‑Powered Content Agency – Offer blog posts, newsletters, and video scripts that are generated, edited & fact‑checked by gpt‑oss’s full chain‑of‑thought reasoning. Use its built‑in function calling to pull real‑time stats or embed charts automatically.



2️⃣ Personalized Study‑Aid SaaS – Build a subscription tool that creates custom quizzes, flashcards, and step‑by‑step explanations for any subject. Fine‑tune the model on curriculum standards to boost relevance.



3️⃣ Niche Market Research Reports – Leverage the model’s agentic web‑browsing (when enabled) to scrape competitor data, trends, and pricing, then deliver concise PDFs to startups that can’t afford big consultancies.



4️⃣ AI‑Assisted Legal Drafting – Provide contracts, NDAs, and policy documents generated under the permissive Apache 2.0 license. The model’s structured‑output capabilities make it easy to produce clean, editable Word files.



5️⃣ Voice‑First Chatbot Builder – Package turnkey voice assistants for e‑commerce sites, using gpt‑oss’s low‑latency inference on a 20 B‑scaled version for quick responses while the 120 B handles complex queries.



6️⃣ Custom Prompt Marketplace – Curate high‑quality prompts for advertising copy, SEO titles, or DALL‑E image generation, then sell them as ready‑to‑use “prompt packs.” The model’s reasoning transparency helps you prove ROI.



7️⃣ AI‑Enhanced Podcast Production – Generate episode outlines, interview questions, and even transcribe recordings with on‑the‑fly summarisation. Use function calls to pull guest bios or recent news for timely content.



8️⃣ Micro‑Learning Video Studio – Create 2‑minute explainer videos that combine AI‑written scripts, AI‑generated visuals, and voice‑over via text‑to‑speech. gpt‑oss’s fine‑tuning lets you match brand tone perfectly.



9️⃣ Automated Data‑Cleaning Service – Offer to clean CSVs, deduplicate records, and enrich datasets using the model’s python tool calls. Small firms love a “clean‑data in 5 min” guarantee.



🔟 AI‑Driven Resume & Cover Letter Builder – Let job‑seekers upload a draft and receive a polished, keyword‑optimized resume plus tailored cover letters, all generated locally for privacy.






Why gpt‑oss:120b?

✅ Agentic: native function calls, web‑search & python tool integration.

🧠 Full chain‑of‑thought: you can surface the model’s reasoning for transparency.

⚙️ Configurable effort (low/medium/high) to balance speed vs. depth.
🎛️ Fine‑tuneable for niche domains or brand voice.
📜 Apache 2.0 license – no copyleft worries, commercial‑ready.
💾 MXFP4 quantisation – runs on 16 GB RAM laptops or a single 80 GB GPU.



Ready to turn your bedroom into a profit hub?
1️⃣ Install Ollama, pull gpt‑oss:120b.
2️⃣ Pick one of the ideas above.
3️⃣ Start building, test locally, and launch globally.



💡 Tip: Offer a free “pilot” using the model’s low‑effort mode to prove value before upselling premium, high‑effort plans.





Let’s hear it! Which of these ideas sparks your curiosity?

Drop a comment or DM – I’m happy to share resources and quick‑start templates. 🌟

4 months ago | [YT] | 1

Hamish from A-Tech

So they finally did it.

we officially have open source LLMs from the company that started it all. (technically speaking of course)


thank you OpenAI.

Definitely a non reasoning one next though please I need real time responses on the daily


Ps, if I don't show up for any meetings today...

this is why 🤣


pps, I'll be live steaming my initial interaction with OSS on YouTube today!

4 months ago | [YT] | 1

Hamish from A-Tech

Does the M1 Max Mac Studio handle AI well?

Find out in this video! 👇

https://youtu.be/XPrYEs6pyGs

4 months ago | [YT] | 0

Hamish from A-Tech

We ordered the Reachy Mini!

More deets coming soon :)

4 months ago | [YT] | 1