Tessera
Folgen

Rainer Hahnekamp

@rainerhahnekamp.bsky.social

AI Engineer at Soverius AI, Google GDE, Trainer at AngularArchitects Co-Maintainer of CopilotKit Angular, NgRx, Sheriff

167 Posts710 Follower147 Folgt
Rainer Hahnekamp@rainerhahnekamp.bsky.social · 9h ago

1/6 Developers are deeply divided over whether human code review is still necessary. Exaggerating slightly: one side accuses the other of abandoning code quality and blindly trusting AI. The counter is that line-by-line review simply throttles AI back down to human speed. 🐢

Rainer Hahnekamp@rainerhahnekamp.bsky.social · last wk.

Not every AI task needs an open-ended answer. Sometimes you just need yes, no, or one of a few options. That’s where Jev comes in. In Quantized AI News 26/08 (by @soverius-ai.bsky.social), we look at its role alongside LLM - and the latest model releases. soverius.ai/blog/quantiz...

Jev, System One AI & New Models | Quantized AI 26/08

Explore TypeSafe’s Jev and System One AI, new models Opus 5.5 and Sol 6.1, and the safety concerns behind Astra 6.1’s cancelled release.

soverius.ai

Rainer Hahnekamp@rainerhahnekamp.bsky.social · 2w ago

We are very much looking forward to this. Agents can be everywhere. We will show Agents - for development, - in the backend, - in the browser, and - model + agent in the browser. Interested? Join our free webinar on Sep. 30 for fresh ideas on what you can "agentize."

Ai Dev Craft@ng-conf.bsky.social · 2w ago

#Agents are becoming part of every layer of the application stack. Join Rainer Hahnekamp and Murat Sari on Sep. 30 for A free #AIDEVCRAFT Webinar and see how #agentic workflows, Mastra, and CopilotKit come together in a modern app. Register 👉 bit.ly/4yXmYLZ

Rainer Hahnekamp@rainerhahnekamp.bsky.social · 2w ago

1/9 Among all that “AI is going to kill us all” talk, it sometimes helps to look at how other high-risk industries are doing it. For example, nuclear power plants. What do they do against a meltdown? Enormous safety efforts in multiple layers.

Rainer Hahnekamp@rainerhahnekamp.bsky.social · 3w ago

I’ll be speaking at #AIDevCraft in Las Vegas! Prompt/Context/Loop/* Engineering There is definitely some hype around these terms, but it’s not only hype. Each phase builds on lessons from the previous one. That evolution is what my talk is about. ai-devcraft.com @ng-conf.bsky.social

Ai Dev Craft 2026 - AI Developer Conference for Engineers

AI developer conference covering generative AI for developers, AI tools for developers, workflows, workshops, and engineering best practices.

ai-devcraft.com

Rainer Hahnekamp@rainerhahnekamp.bsky.social · 4w ago

1/10 After giving my talk on local LLMs for software development three times, here are some recurring questions and assumptions I've heard so far: “I have a subscription. I don’t need local models.” That works as long as you stay within your allowance. With manual prompting, that might be enough.

Rainer Hahnekamp@rainerhahnekamp.bsky.social · last mo.

Switzerland is an Angular country 🇨🇭 57% of Swiss frontend jobs naming a framework ask for #Angular. It’s time for a Swiss Angular conference. It’s happening and @wolfmanfx.bsky.social and I (both @soverius-ai.bsky.social) are there with workshop & talk: ng-switzerland.ch

NG Switzerland — 11–12 February 2027, Lucerne, Switzerland

The Angular conference in the heart of Switzerland. 11–12 February 2027, Lucerne, Switzerland.

ng-switzerland.ch

Rainer Hahnekamp@rainerhahnekamp.bsky.social · 3mo ago

Many users complain that Fable falls back to Opus 4.8 too quickly because of strict guardrails. We need guardrails. But what happens when safety makes a model feel worse, especially if other models answer more freely? That is the dilemma in this week’s Quantized AI News soverius.ai/blog/quantiz...

Quantized AI News 26/05: The Price of Control in Frontier AI

Frontier AI is no longer just about capability. This week: Fable’s guardrails, cheaper tokens, and governments moving closer to AI infrastructure.

soverius.ai

Rainer Hahnekamp@rainerhahnekamp.bsky.social · 3mo ago

Fable 5 will be available again. A lot of people look at the Fable story and see the U.S. government as the bad actor that blocks technological progress by restricting access. But that is too simple. 1/5 soverius.ai/blog/quantiz...

Quantzied News: GPT-5.6, Claude Tag, and the Distillation Problem

Recent AI updates (GPT-5.6, Claude Tag, Alibaba) shift focus from raw power to access and control: who uses these models, how, and where their outputs go.

soverius.ai

Rainer Hahnekamp@rainerhahnekamp.bsky.social · 4mo ago

The latest edition of Quantized AI News (by @soverius-ai.bsky.social) is out. My main highlight is the chapter about Cognitive Coverage: the ongoing debate about whether we should still look at the code. soverius.ai/blog/quanitz...

Fable Shutdown, GLM-5.2 & OpenRouter Fusion, Cognitive Coverage

State AI control is rising: the US shut Fable 5 but kept xAI running. A Chinese open-weight model offers an alternative. Plus: AI and Cognitive Coverage.

soverius.ai

Rainer Hahnekamp@rainerhahnekamp.bsky.social · 4mo ago

The fastest way to understand modern AI is to run a model inside your own, self-written harness. One hour. Local machine. No mystery. Here's the step-by-step article by @wolfmanfx.bsky.social for @soverius-ai.bsky.social. soverius.ai/blog/impleme...

Build a local AI coding agent from scratch

Build a local AI coding agent from scratch — Gemma 4 on llama.cpp, three tools, one loop — then learn why running it unsandboxed is dangerous and how NVIDIA OpenShell contains it.

soverius.ai

Soverius AI@soverius-ai.bsky.social · 4mo ago

Fable 5 is a good reminder why AI sovereignty matters. It is powerful, but access, pricing, and availability can change quickly. Local LLMs will lag behind the frontier, but you get 💰 Cost control 🔐 Data privacy 🏠 Independence Interested? Register 👇 soverius.ai/webinars/sof...

Software Development with Local LLMs

It is time. Token costs rise constantly. Are local LLMs ready to take over? Join our free webinar and see for yourself. In part one, Rainer shows the setup of a local LLM. He connects it to a modern ...

soverius.ai

Rainer Hahnekamp@rainerhahnekamp.bsky.social · 5mo ago

As teased last week, my article about harnesses has finally been published. This is something I wanted to write for a long time. Not just what it is, but how the term evolved and why it is so important for AI in general. Big thanks to @wolfmanfx.bsky.social for the review & discussions.

Soverius AI@soverius-ai.bsky.social · 5mo ago

For a long time, AI discussions were mostly about the model. 🤖 Which one is better? 📊 Which benchmark was beaten? 🚀 Which new release changed everything? But only the model is not enough. The layer around it is just as important That's the harness. 📖 soverius.ai/blog/what-is...