shreezus

Anthropic is one of the few companies that will likely hit an infinite valuation if they succeed in their mission.

Unfortunately, there is almost no point buying this when the MacBook Neo exists, and runs a full-fledged operating system rather than ChromeOS or Gemini or whatever it is they’re calling it.

I have personally witnessed AI capabilities that would be considered “impossible” by current public standards.

The next 2-3 years are going to be incredibly interesting.

AGI doesn’t need to be “solved” for humanoid robots to be valuable at scale. The role of teleoperation is often underestimated; in the near term, many humanoids will likely be operated remotely by people halfway across the world, performing deliveries and other tasks cheaply.

This is a real risk. I know of someone who got phished with a fake number for Apple Support (the fake number was promoted and appeared at the top of the search results). Apparently they do this with banking phone numbers as well.

Spotify needs to embrace a user-centric payment model (where a portion of your subscription fee goes directly toward artists you listen to the most) rather than the current “pooled” system which massively favors large artists.

YouTube, for instance splits 55% of its YouTube Premium subscription revenue with creators, who get paid based on their share of watch time. Your YouTube Premium fee is distributed to the creators you view the most.

University research grants that have the word "mRNA" present are currently being flagged and frozen, even though mRNA technology has been used for things like cancer vaccine research for years. Politicizing a technology is incredibly absurd and will have long-term repercussions on science & medicine.

I know of a professor at one university that had grants frozen due to being flagged as "woke" gender discourse. His lab researches...(wait for it)... immunotherapy treatments for breast cancer in women.

There are a lot of "interesting" emergent behaviors that happen just a result of scaling.

Kind of like how a single neuron doesn't do much, but connect 100 billion of them and well...

I own a Tesla, and here's my take on the biggest software issue:

Normal consumers don't understand the difference between "Autopilot" and "FSD".

FSD will stop at intersections/lights etc - Autopilot is basically just cruise control and should generally only be used on highways.

They're activated in the same manner (FSD replaces Autopilot if you pay for the upgrade or $99/month subscription), and again for "normal" consumers it's not always entirely clear.

A friend of mine rented a Tesla recently and was in for a surprise when the vehicle did not automatically stop at intersections on Autopilot. He said the previous one he rented had FSD enabled, and he didn't understand the difference.

IMO Tesla just needs to phase out 2019 AP entirely and just give everyone some version of FSD (even if it's limited), or geofence AP to highways only.

I wonder if this is related to the capacity/uptime issues Anthropic has had lately. I got quite a lot of errors last week!

Hopefully they sort it out and increase limits soon. Claude Code has been a game-changer for me and has quickly become a staple of my daily workflows.

LinkedIn is probably the worst culprit. It has always been a wasteland of “corporate/professional slop”, except now the interface deliberately suggests AI-generated responses to posts. I genuinely cannot think of a worse “social network” than that hell hole.

“Very insightful! Truly a masterclass in turning everyday professional rituals into transformative personal branding opportunities. Your ability to synergize authenticity, thought leadership, and self-congratulation is unparalleled.”

I believe we will see integrated optical glucose sensors in a popular consumer wearable fairly soon.

I don't think they'll be as accurate as blood sensors, however they will be a game-changer for many people (pre-diabetics, or gestational diabetes etc).

As someone who has been at a company where for various reasons, we decided to "roll our own auth", I would have to disagree here. Don't reinvent the wheel if you can avoid doing so.

I mostly agree with you. I use both FSD on my Tesla & Waymo regularly (LA region), and Waymo just feels way safer in comparison. While FSD has improved significantly the last few iterations, I have seen it do "strange" things often enough that I don't feel safe just sitting in the backseat like I would with a Waymo.

Even if it's hypothetically 99% as good as Waymo at the moment, 99% is not good "enough" when it comes to something as critical as driving.

Agreed - I disabled all non-essential notifications (I don't need Slack pinging my wrist) and have found my watch actually helps me ditch the phone more easily.

I'm still "reachable", but the watch UX is annoying enough that I won't find myself scrolling X etc on it.

Is anyone else totally blown away by this? I feel like it’s easily the biggest announcement out of IO, however it’s been overshadowed by Veo 3 etc.

Diffusion models for code generation are a big deal. If they are using transformers this would likely fall into the DiT bucket (diffusion transformers). I had previously worked on use cases that leveraged U-Net diffusion several years ago and there was quite a bit of interest in hybrid models. I expect to see further leaps in the diffusion space in the near future.

Exactly - I have tinkered with all sorts of systems (including Obsidian), and ultimately prefer the simplicity of Apple Notes. I have a system of nested folders and tags that handles my organization needs pretty well, and I find added complexity just ends up adding more friction to the note-taking process. Apple Notes search could certainly use improvement though.

I have recently been impressed by the quality of Mexican coffee (currently drinking one grown in Chiapas). Medium roasts tend to play nicely with beans from this region.

It's a bit further than that, the service area now extends to DTLA / Arts District and south toward Inglewood (likely to support LAX soon).

It's one of those things that's impressive initially, but after generating a couple it feels quite formulaic.

This is only a problem for non-Teslas or vehicles that can't utilize the Tesla supercharger network.

I don't think I have ever plugged into a supercharger more than 10/15 minutes.

Up until recently (~2022/23) Toyota had cumulatively sold more hybrids than all EVs sold by all manufacturers combined, globally. They arguably have the best hybrid drivetrain on the market, and it's gotten to the point where even the Camry (2025 onwards) are exclusively offered as hybrids now.

I have noticed the same - Tesla FSD has improved remarkably in the last couple of versions, to the point where it almost never requires interventions.

That said, I still don't "trust" FSD to the point where I could comfortably be in the backseat taking a nap just yet. 99% is not "good enough" for something as critical as driving.

There are so many interesting, off-label novel uses for drugs that have already been developed. Recently my wife & I adopted a kitten that developed FIP, a disease caused by a feline coronavirus that was nearly 100% fatal up until just a few years ago.

Turns out, a drug developed by Gilead Sciences (GS-441524) which is the active metabolite of remdesivir, is remarkably effective at treating the disease. It's a nucleoside analog, so it essentially is a slightly modified adenosine molecule which disrupts the virus RNA replication process.

Within 12 hours of the first dose our cat went from dying to making a full turnaround and complete recovery following treatment. For us it was truly a miracle drug, and it only gained awareness recently due to the COVID-19 pandemic [1]. It still isn't technically "approved" for this use case, but is prescribed off-label by vets as of 2024 once the FDA relaxed its position [2]. Folks needed to seek it out on the black market (and still do in some areas).

[1] https://www.theatlantic.com/science/archive/2020/05/remdesiv...

[2] https://www.fda.gov/animal-veterinary/cvm-updates/fda-announ...

Haven't had a chance to play with this yet, but 10M context window is seriously impressive. I think we'll see models with 100M context relatively soon, and eliminate the need for RAG for a lot of use cases.

Even if you haven't personally used their service, if any close relatives have, they already have a sizable amount of information on your genome. They maintain the equivalent of "shadow profiles" (https://en.wikipedia.org/wiki/Shadow_profile) as part of their data model for "ancestry" modeling purposes - for example inferring a paternal haplogroup based on data uploaded by genetic relatives.

I can only hope at the end of the day their data doesn't end up in the wrong hands. It is their most valuable asset, and this is a way bigger deal than it seems.

This is why agentic AI will likely cause a cataclysim in white-collar labor soon. The reality is, a lot of jobs just need "OK" performers, not excellent ones, and the tipping point will be when the average AI is more useful than the average human.

Both can be true at the same time.

That said, utility companies have a duty to properly maintain infrastructure, and this is an easier problem to fix. There are a few companies working on this, such as Gridware (https://www.gridware.io).

This is why we need a push for radical transparency when it comes to public spending.

What did they spend $556,539 per inmate on, exactly?

Taxpayers deserve an itemized receipt.

Otherwise, by accepting overspending as the norm, it obfuscates corruption & large-scale grift as taxpayers get fleeced.

There's also substantial opportunity cost - that's $500K+ that could be going toward education or useful public services, that ended up in someone's pockets instead.

Transparency invites accountability.

Well-deserved. Congratulations on the exit, and looking forward to seeing what you build next!