Mihaly Hanics @HanicsResearch

my parents' old Android TV was full of bloatware and sponsored rows you couldn't turn off. so I had Claude rip it all out. Claude Code ran on my Mac mini and did everything over ADB. it disabled the bloatware, the tracking and the auto-installers, and uninstalled 11 apps nobody used. free RAM went up about 90%. then it built a custom launcher with just the apps they actually use. and every change can be undone with one command. honestly it feels like a new TV now.

Ryan Doyle @doooyle

4 days ago

This is how fast the TV moves now that Claude deleted all the BS laggy bloatware it comes with

348 831 27K 3.0M 12K

@psviderski In Rust the traditional glyph is an amongus.

With the MI355X generation and great software improvements, AMD is now NVIDIA tier. And with NVIDIA pulling the 5090, their consumer GPUs are competitive too. It's a damn shame Intel doesn't have people to antagonize on Twitter. They could be next, but alas, no leadership.

OpenCode found their niche Smart models too cheap to meter

OpenCode @opencode

5 days ago

GLM 5.3 Flash is now free in OpenCode

788 1K 20K 2.8M 5K

Some observations about the "total" automation of software engineering: 1. There are more software engineers now than in 2025, and demand is growing. 2. Software engineers have never worked harder than they do now. 3. They largely don't write code anymore, agents write ~100% of the code. 4. The bottom line of the companies that employs these engineers and agents (i.e. the terminal purpose of all this code) is typically not budging as a result of AI adoption. This is why you have to think in terms of systems. Automating one aspect of a system creates its own friction and moves the bottlenecks elsewhere.

You can do so much for free on whatever device you own

Youssof Al Toukhi @Youssofal_

7 days ago

This is very true. Here is Qwen 3.8 27B compared to all the sota models 1 year ago today on artificial analysis. If 27B released a year ago, it would be the best model in the world by a wide margin.

54 128 2K 239K 497

Incredible: Not only is Google working to block AI crawlers, but @nozzleio uncovered evidence that Google is deliberately serving FAKE search results to scrapers The "poisoned" search results look real, but w/ garbage data. This influences AI crawlers, scrapers, & rank trackers.

Definitely disagree with this, and moreso the higher you get. Lots of intellectually demanding jobs and fields of research that attract/require high IQs don't pay commensurately. Even in STEM a truly brilliant scientist or engineer doesn't necessarily discover something revolutionary, or his discoveries are more abstract or not immediately applicable, or made as part of a larger group, and/or are not commercialized at all or in a way that benefits him specifically. Or some combination. Not that they won't do fine financially, but their income percentile is likely not matching their IQ percentile in the vast majority of cases, for most if not all of their working lives. Just as an example, a top 1% IQ ("only" 135) in the US would have to earn 560k-800k. Even top doctors, lawyers, and engineers usually earn significantly less than that and those that do earn more often work insane hours and deal with high amounts of stress. Not everyone is cut out to be an entrepreneur/business owner, let alone a scalable business, and that's true even among the very intelligent. And that's in a functional meritocratic system, not the Affirmative Action/DEI/woke/HR/etc dystopia we've been experiencing for over 50 years. Add that nonsense on top and it looks even worse.

تخيل تربط Apple Watch بأندرويد بدل ما تكون محصوره بالآيفون مشروع مفتوح المصدر يحاول يكسر هذا الاحتكار، ويوفر اتصال الساعة بأجهزة وأنظمة مختلفة، مع إمكانية إدارة واجهاتها وبعض وظائفها. • التجربة الفعلية مو مكتملة لكل الأجهزة، ونسخة أندرويد تحتاج Root وMagisk. تم اختبار المشروع على Apple Watch Ultra 2 بنظام watchOS 26.2.

@thekitze Kind of lost track, so this is yet another personal agent? What happened to Hermes and OpenClaw, you used to have like hundreds of claw pipelines? The new alternatives are better designed, or? (I recall people complained OpenClaw is buggy, but I assume it got many fixes..)

@badlogicgames @mitsuhiko @steipete It'd be fun if the concept was solving complex problems, as in theoretically complex. Some would work on scientific problems, some would try to find math hypotheses proofs, some would try to find engineering solutions to problems unsolved, so on. Unsolved challenging problems

finally landed on SF, but we might need to go back to Brazil :( long story short, I was really tired from the trip, hospital and the migraines, and I wasn't at all prepared for all these questions. the immigration agent didn't love my English pronunciation nor my story, asked tons of questions and decided to revoke our visas, giving us 2 weeks to go back this is kinda devastating to HOC because we had so many meetings lined up and even some soft commitments. also we had already paid for it all... but yeah, my bad, next time I'll be prepared. I think I'll just refund HOC, reschedule the conversations to calls, and use the time here to focus on the health checks? there may still be ways to recourse though, I've spawned a fleet of agents to investigate my options on a brighter note I'm finally home, safe, no migraines, and the food here is amazing. everything seems so high quality. although I'm concerned about the french fries, are you sure you're not confusing salt with sugar? I'm not complaining, it tastes amazing, but that cannot be healthy?

The future depends on the direction this trend takes.

@_modprobe @itsmattchan @mattpocockuk The worst is when a team is full of tornadoes ("pirates") and they take one guy to coordinate the mess left around.

@_modprobe @lennysan @danshipper I agree, the architect hates "value-finding" (or credit-taking) pirate who does more harm around than value alone. This role existed far before agentic coding, the guy who would never take a moment to design a system well and just get straight to writing code without a good map.

@mattpocockuk I ran /ask-matt and seemingly, it doesn't know about updates: "No, the current v1.3 skills do not appear to know their migration history." Perhaps consider adding the updates as reference

@mattpocockuk Matt, the GLOSSARY change made me feel the need for an "update repos based on skills updates" skill for those who ran /setup-matt-pocock-skills in a repo earlier: e.g. if it was set up with v1.2, it has CONTEXT.md files. Calling this skill would do renaming. Or /ask-matt does it?

Meta removed human support from Instagram, it’s AI support only for paying customers. Which is a decision, but OK. But the AI support doesn’t work. As in, at all. So I pay for a product with no support, no way to report this, and no way to get a receipt event FML

In my view, the right response by the mathematical community to OpenAI’s papers is to say: “We’re not accepting these papers as finished work. It’s not our job to sit down and read your model-generated proofs. It is your job to present your work through presentations and convince people of the correctness of your work. It’s always the case that whoever makes a claim needs to provide sufficient explanation and evidence to make independent verification feasible. Until that moment, you just have a claim.” There’s a difference between a proof being mathematically correct and a proof being accepted as correct by the mathematical community. Mathematical truth doesn't depend on consensus, but scientific recognition requires scrutiny, communication, and reproducibility. They're still the arbiters of whether anyone's claims, including OpenAI's, are accepted as valid, and they should leverage that to the maximum extent possible.

@mattpocockuk @asmirkn @poteto and you will have a working, but rotten solution. As the service advances, always having AI deal with it end-to-end will have no issue, but a script will turn into being outdated quickly as everything around it updates

@mattpocockuk @asmirkn @poteto API changed, website being scraped changed, data model changed, pipeline changed, something new was added that you never noticed - because your script offers the agent only processed data and goes unnoticed. It's not something that loudly fails, because of backwards compatibility