Blog

Posing 80.000 en beter werkt t deze x wel XD. -_-

A lot of people have been sleeping a little less lately. I know I have. But the potential upside of getting this right (along with the fun of building at this edge) makes the shorter nights worth it.

Over the past year, like so many, we’ve steadily replaced core workflows at Base with AI systems built in-house. That matters to us because control over the system means control over cost, security, and capability.

In January, we shifted to building with ClawdBot. Open-source, result-driven execution, and persistence gave us flexibility and power packaged into one. It also meant not being locked into someone else’s roadmap or pricing model. But experimental technology often comes with security risks. And so did OpenClaw.

So we had to tackle security first
Then invest in infrastructure
Optimized token usage to reduce cost
Expand infrastructure
Benchmarked fully local models against hybrid performance
Also bought toys and explored fun use cases 😊

At a certain point, it became clear this isn’t just a productivity tool. It is a platform layer, capable of supporting real business operations. And as far as we’re concerned, the AGI debate can wait. There is still more than enough value to be captured with secure, specialized AI that solve real business problems.

So what are we trying to get right?

Today, AI covers roughly 90 percent of nearly one-third of Base’s internal workflows. In most cases, with better consistency than manual execution. Our aim is for AI to handle 100 percent of our 𝒃𝒖𝒔𝒚𝒘𝒐𝒓𝒌; predictable, repetitive work that takes any team member longer than 15 minutes.

We intent to achieve this…
Not by hiring more people
Not by increasing overhead
Not by adding another SaaS subscription

We want to do this in-house, with sovereign systems we fully control, at cost neutral or better. Automation only wins if it is secure, sustainable and makes financial sense.

• On this journey, 𝐨𝐮𝐫 𝐭𝐞𝐚𝐦 𝐢𝐬 𝐬𝐭𝐢𝐥𝐥 𝐥𝐨𝐨𝐤𝐢𝐧𝐠 𝐟𝐨𝐫 𝐨𝐧𝐞 (𝐨𝐫 𝐭𝐰𝐨) 𝐦𝐨𝐭𝐢𝐯𝐚𝐭𝐞𝐝 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐞𝐫 𝐢𝐧𝐭𝐞𝐫𝐧𝐬 𝐭𝐨 𝐣𝐨𝐢𝐧 𝐮𝐬 until the summer (for info on a similar role see: https://lnkd.in/ecp5TFGJ) If you want to work on real operational AI systems, not just experiments (but definitely 𝐚𝐥𝐬𝐨 𝐞𝐱𝐩𝐞𝐫𝐢𝐦𝐞𝐧𝐭𝐬 𝐥𝐢𝐤𝐞 𝐭𝐡𝐢𝐬 𝐎𝐩𝐞𝐧𝐂𝐥𝐚𝐰 𝐨𝐧 𝐚 𝐌𝐚𝐜 𝐦𝐢𝐧𝐢 𝐚𝐧𝐝 𝐌𝐚𝐜 𝐔𝐥𝐭𝐫𝐚 𝐰𝐢𝐭𝐡 𝐑𝐀𝐘-𝐁𝐀𝐍 𝐌𝐞𝐭𝐚 𝐀𝐈 𝐠𝐥𝐚𝐬𝐬𝐞𝐬!) – reach out. Or please pass this along to someone who might be a fit.

If you are building practical AI systems inside your own organization, I would genuinely enjoy hearing your stories and comparing notes. What worked, what failed, what surprised you. We are always learning and are interested to find out what we are overlooking. We’re all doing our best trying to figure things out in real-time 😊…….

hdbcgjhdbccjkdbjkbcJKbc bjbjbjUIDJKDJKDJKsfsfsfsfsfsfsfsfsfsfsfsfsfsfsfsfl>

A lot of people have been sleeping a little less lately. I know I have. But the potential upside of getting this right (along with the fun of building at this edge) makes the shorter nights worth it.

Over the past year, like so many, we’ve steadily replaced core workflows at Base with AI systems built in-house. That matters to us because control over the system means control over cost, security, and capability.

In January, we shifted to building with ClawdBot. Open-source, result-driven execution, and persistence gave us flexibility and power packaged into one. It also meant not being locked into someone else’s roadmap or pricing model. But experimental technology often comes with security risks. And so did OpenClaw.

So we had to tackle security first
Then invest in infrastructure
Optimized token usage to reduce cost
Expand infrastructure
Benchmarked fully local models against hybrid performance
Also bought toys and explored fun use cases 😊

At a certain point, it became clear this isn’t just a productivity tool. It is a platform layer, capable of supporting real business operations. And as far as we’re concerned, the AGI debate can wait. There is still more than enough value to be captured with secure, specialized AI that solve real business problems.

So what are we trying to get right?

Today, AI covers roughly 90 percent of nearly one-third of Base’s internal workflows. In most cases, with better consistency than manual execution. Our aim is for AI to handle 100 percent of our 𝒃𝒖𝒔𝒚𝒘𝒐𝒓𝒌; predictable, repetitive work that takes any team member longer than 15 minutes.

We intent to achieve this…
Not by hiring more people
Not by increasing overhead
Not by adding another SaaS subscription

We want to do this in-house, with sovereign systems we fully control, at cost neutral or better. Automation only wins if it is secure, sustainable and makes financial sense.

• On this journey, 𝐨𝐮𝐫 𝐭𝐞𝐚𝐦 𝐢𝐬 𝐬𝐭𝐢𝐥𝐥 𝐥𝐨𝐨𝐤𝐢𝐧𝐠 𝐟𝐨𝐫 𝐨𝐧𝐞 (𝐨𝐫 𝐭𝐰𝐨) 𝐦𝐨𝐭𝐢𝐯𝐚𝐭𝐞𝐝 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐞𝐫 𝐢𝐧𝐭𝐞𝐫𝐧𝐬 𝐭𝐨 𝐣𝐨𝐢𝐧 𝐮𝐬 until the summer (for info on a similar role see: https://lnkd.in/ecp5TFGJ) If you want to work on real operational AI systems, not just experiments (but definitely 𝐚𝐥𝐬𝐨 𝐞𝐱𝐩𝐞𝐫𝐢𝐦𝐞𝐧𝐭𝐬 𝐥𝐢𝐤𝐞 𝐭𝐡𝐢𝐬 𝐎𝐩𝐞𝐧𝐂𝐥𝐚𝐰 𝐨𝐧 𝐚 𝐌𝐚𝐜 𝐦𝐢𝐧𝐢 𝐚𝐧𝐝 𝐌𝐚𝐜 𝐔𝐥𝐭𝐫𝐚 𝐰𝐢𝐭𝐡 𝐑𝐀𝐘-𝐁𝐀𝐍 𝐌𝐞𝐭𝐚 𝐀𝐈 𝐠𝐥𝐚𝐬𝐬𝐞𝐬!) – reach out. Or please pass this along to someone who might be a fit.

If you are building practical AI systems inside your own organization, I would genuinely enjoy hearing your stories and comparing notes. What worked, what failed, what surprised you. We are always learning and are interested to find out what we are overlooking. We’re all doing our best trying to figure things out in real-time 😊

New Post: [POST_TITLE] – [POST_LINK][POST_EXCERPT][POST_CONTENT][POST_AUTHOR][WEBSITE_TITLE]

test text

test text 2

Welcome to WordPress. This is your first post. Edit or delete it, then start writing!