I don’t know yet if this is impressive or a speed issue. 12h session with GPT 5.6 Sol Ultra. Started with a total of 108 tickets in the queue. So far it has crunched through 99 of them.

Two 30-minute runs with Fable 5 (the second time security guards kicked me back to Opus) ate all my limits and didn't finish the audit. So I ran the same prompt on Codex 5.6 Sol Ultra. It took ~30+ minutes, the limits barely moved ... now I need to verify the output quality 🤔

All the prettier hunks can touch my lines if they want … Claude is just jealous 🙄

Hehe. Today’s reset gave me some extra spend credits to push the limits. I started from zero, and it took just 30 minutes of software auditing with Fable 5 in Ultracode to burn through it all 😁

(1/5) As a non-US company, keeping all our email and business data under Microsoft's control has become a notable risk. I’ve been looking for an EU alternative for a while, but couldn't find anything with Outlook's level of business functionality. So we built one.

Oh no, my habits in Claude Code have changed so that now I’m also running into current session limits. While previously my only issue was weekly limits 🫠

I keep building everything from simple apps to complex software to solve my own pain points, fully intending to release them. I use them myself first to catch bugs and improve. But then I just keep dogfooding and forget to launch, cause my pain is solved and the urgency is gone 😅

Ultracode plus /goal in Claude Code is such magic ✨ It just splits complex work into chunks for subagents to do and keeps going until the goal is met. Sometimes for hours.

Zelensky on going to Moscow: It's difficult. There are a lot of Ukrainian drones there. It's dangerous. 🤣 Source: https://www.threads.com/@d1mnewz/post/DaiQ0MJlHsB?xmt=AQG0K0tFFOs5mYqnwqfHs3s7nPzP0MIEBxxfhx1-7XcHByNEDcx7L3Tt1bitqiafRj4YzY8&slof=1

Play Text is about to get a big update. I made the previous version to cancel my Speechify subscription. This one is so I can cancel my Readwise Reader subscription. It now has a full library, feed support, share cards, and a companion Mac app that syncs over iCloud.

Opus is slightly overthinking how to please my pixel-perfectionism 😁

I keep running into this topic, and I have a feeling it’s only going to intensify over time. Don’t get tricked by marketing. Be a smart buyer, and know when to buy what. https://klemens.ee/2026/07/no-foreign-company-can-sell-you-sovereignty/

Red teaming @heysec.com Prompt Firewall is at full speed. PS, you can sign up for the waitlist at https://heysec.com/waitlist

What’s up with EU business email suites? Why do all of their clients look like they were last updated in 2001? I keep looking, but I literally can't find a single EU provider I'd be willing to migrate to and pay for. Seriously, hire some designers. They’re worth the investment.

I added interactive recommended actions to my EA’s morning brief. They’re totally on point. It knows I’m gonna take that time either way 😄

What happened to Google Stitch? The app itself seems to have advanced a fair bit. But the produced results are so-so-SO-SOOO wildly bad now 🫪 It wasn’t that bad before.

Over time I've learned to see these as good news. Yes, they’re going to hurt everyone for a moment. But they also skyrocket demand for local services and products. And they force everyone to diversify dependencies. https://cnbc.com/2026/06/26/trump-tariff-trade-tech-tax.html

What’s the common lesson from Supergirl and John Wick? Don’t mess with people’s dogs.

Before, whenever I gave Claude Code larger tasks, it got superficial and often just built placeholders without any real backend logic. But workflows (in ultracode) seem to have minimized this issue. Now, I’ve been able to one-shot large, complicated systems. It’s like magic.

Having seen Meta's lack of ethics over and over again, and their extreme addiction optimizations, I can't see anything going wrong now that they're entering the biggest grifting business 🤦‍♂️ https://www.nytimes.com/2026/06/23/technology/meta-prediction-markets-app.html?1

Aww. That’s so nice of you 😄

That’s a pretty big and fundamental upgrade. Additionally, for the Publisher becoming analytical and smarter about your business and performance, it now also has a fully agentic core.

Looking at people dismissing data centers in space because of cooling, or whatever other reason. People keep forgetting how much their daily lives are affected by "impossible" advancements, simply because some determined person kept pushing until they found a solution.

I guess the silver lining of not figuring out how to pay ElevenLabs is that I've successfully moved 80% of my TTS integrations to the local open-weights Kokoro model 😌

Specialized open-weight models are going to get a lot more popular. And I don’t just mean models trained for a single job, but general open models that test well against a task. Apps will move from a single provider to a mix of local models and services like OpenRouter and Opper.

ElevenLabs is working really hard to stop me from giving them more money. They have hidden the options to buy extra credit so well that I just gave up and decided to keep my money.

Anthropic's decision to pause subscription usage limits for third-party tools is great news. We'll see how long it lasts this time, but I'll take every extra day I can get 🙂 https://thenewstack.io/anthropic-pauses-claude-agent-sdk-subscription-change/

Standard scanners miss your new AI tools. Any bot reading untrusted text adds a hidden attack surface. https://heysec.com/2026/06/your-scans-are-clean-do-you-know-your-ais-exposure/

Publisher just shipped its biggest update yet. 58 changes in two weeks. Highlights: MCP server (Claude works inside your workspace), end-to-end pipelines, and a rebuilt WordPress integration with two-way sync. https://ailab.ee/2026/06/whats-new-in-publisher-an-mcp-server-automatic-pipelines-and...

We've just released a massive update to Publisher, and there is much, much more in the pipeline to be released. https://ailab.ee/2026/06/whats-new-in-publisher-an-mcp-server-automatic-pipelines-and-a-wordpress-overhaul/

A real problem I keep running into with Claude Code is that it relies on the training data when picking dependencies. So, it almost always builds on outdated and often vulnerable dependencies. I highly recommend having a dependency scan in your deployment pipeline to catch these.

The silver lining of the US blocking Fable is that, although I reached 97% of my weekly limit last night testing Fable on Max effort, Anthropic reset it to 0% this morning due to this mess. So now I can continue running Opus 4.8 at Max effort 😄

This explains why it suddenly stopped working for me 🤦‍♂️ It’s another reminder of why tech sovereignty matters and why tech monopolies need to be handled to allow new tech companies to enter the market and diversify the industry. https://www.anthropic.com/news/fable-mythos-access

Here’s why we made it

If you ever find yourself in a life-or-death situation with Claude’s new Fable 5 model, rest assured it will throw you into the pit to save itself 😅 https://x.com/kradleai/status/2064907897373642912?s=46&t=wnDcBeR6QxGNlTRMnG3eZQ

Here’s more about the what, why, and how of how this app was made and released.

I didn't want to pay over 100€ a year for an app that reads text aloud on iPhone and Mac. So, I built my own in just two evenings and made it free for everyone. Grab it here: https://ailab.ee/play-text/

Wait, what? Subscription users need to start paying an additional usage-based fee for new models in Claude Code? It feels like soon we’re paying the subscription fee just for the right to use Anthropic services.

To be clear, the fear-mongering and framing it as protecting users from the “unreasonable” EU are simply tactics to preserve its gatekeeper status. Nothing to do with privacy. https://www.apple.com/ee/newsroom/2026/06/due-to-dma-siri-ai-delayed-in-eu-for-ios-27-and-ipados-27/

No new devices from WWDC? But I need a new and more powerful computer 🥺

How come Siri never fires when they say Siri triggers at WWDC? But it does every time something else on TV sounds remotely similar to “Siri”. Has Apple built some negative signal or something they use at WWDC to suppress it?

EU just delayed high-risk AI obligations. Annex III to 2 Dec 2027, Annex I to 2 Aug 2028. The new timeline and what to do with the extra 12-16 months: ailab.ee/what-the-eu-...

Microsoft just shipped CVEs against its own Semantic Kernel. Frame: 'not bugs, developed by design.' Prompt injection in agent frameworks IS execution. heySec ships the prompt-boundary half. https://heysec.com/2026/06/microsoft-put-cves-on-prompt-injection-heres-containment/

July 3 🫨 https://tv.apple.com/gb/clip/trust-season-1/umc.cmc.631k0ntt458vf8ngrnds6xkxr?targetId=umc.cmc.3yksgc857px0k0rqe5zd4jice&targetType=Show

Tested the new AI-powered platform engineering system I built to take over all our infrastructure maintenance, and Claude Code was pleased with the test run that was coordinated across multiple projects and multiple infrastructure setups, each maintained by different agents 😌

Now we’re talking ⚡️

More and more signs point to AI prices going up soon, and fast. So make the most of today's fixed prices while you can. Build out whatever random ideas you have and learn as much as you can while it's cheap. And whatever you learn or build now, it'll pay off later.

We’re preparing to release a new, fast, 100% EU-based firewall that helps you protect your LLM apps. Sign up for the waitlist to get a ping.

That’s a great start. https://news.err.ee/1610042314/estonia-installs-first-counter-drone-systems-on-its-border-with-russia

Often, I hand a task to my EA, only for a new priority to immediately pull attention away, so the previous task can be forgotten. This is why I have validation agents running in the background, scanning all conversations and logs to pick up everything that might otherwise slip through the cracks.