Sitemap
Igor Costa

Igor Costa

Software Craftsmanship, Founder and CEO at Autohand.ai

Open intelligence

2 min readJun 10, 2026

Today Anthropic announced Claude Fable 5.

Most people are talking about benchmarks, context windows, and coding performance.

What caught my attention was something else. (Which now they removed from the announcement if you want to read it again.

Press enter or click to view image in full size

In their system card, Anthropic explains that they have introduced safeguards that intentionally reduce Claude’s effectiveness on certain frontier AI development tasks, including areas such as pretraining pipelines, distributed training infrastructure, and ML accelerator design.

More importantly, these interventions are designed to be invisible to the user.

I understand why they are doing this.

Building increasingly capable AI systems raises real questions about safety, competition, and the pace of technological progress.

But this announcement highlights a fundamental philosophical difference that is emerging across our industry.

At Autohand, we believe intelligence should be transparent, distributed, private, abundant, and grow with you.

Not rented.

Not rationed.

Not silently restricted.

If a model refuses a request, users should know.

If a system has limitations, users should understand them.

If safeguards exist, they should be visible and explainable.

The moment intelligence begins changing its behaviour without the user’s knowledge, we move away from tools and closer to gatekeepers.

That is not the future we want to build.

The history of technology has been a story of expanding access.

The printing press expanded access to knowledge.

Open source expanded access to software.

The internet expanded access to information.

AI should expand access to intelligence.

Get Igor Costa’s stories in your inbox

Join Medium for free to get updates from this writer.

Not concentrate it.

Not permission it.

Not hide it behind invisible controls.

This belief is why Autohand is built in the open.

Our projects are open source.

Our tooling is open.

Our research is published.

Our users choose the models they want to run.

Our goal is not to decide who gets intelligence.

Our goal is to make intelligence available to everyone.

We believe intelligence should be distributed like a forest.

No single tree controls the ecosystem.

Knowledge flows through the network.

The system becomes stronger as more participants contribute.

Resilient systems are not built through central control.

They emerge through openness, collaboration, and transparency.

The future we see is one where intelligence is personal, private, and owned by the people who use it.

A future where memory grows with you.

A future where models learn alongside you.

A future where intelligence is abundant.

That future requires trust.

Trust requires transparency.

And transparency requires openness.

The future of AI is not permissioned intelligence.

The future of AI is open intelligence.

Igor Costa
Founder & CEO
Autohand AI

--

--

Igor Costa
Igor Costa

Written by Igor Costa

Software Craftsmanship, Founder and CEO at Autohand.ai