I trained myself to take a power nap to track “2/2” on that record. After a really hard workout, getting cleaned up, having a bite to eat and drifting off for 30 minutes so many times it has a warm spot in my heart. I like that music can serve as “functional”. I also recommend Mindspring Memories if you have nostalgia of the personal computing boom for a unique ambient. To me it evokes what trying out a Phillips CD-i at my local department store felt like in 1991. YMMV
I wonder if the comments will demonstrate responses that often reference an effect, theory, law, truism, named phenomenon, or some other thing that people excellent at pattern recognition would surface to explain or model the topic at hand. “What you’re describing is Jevon’s Paradox.”
With vinyl warmth is the result of a deliberate process. Professional masters are done specifically for vinyl to accommodate its quirks which truly changes the sound. They have to clamp down the dynamic range and tidy low frequencies or the needle will skip. Recordings with lots of busy high frequency information also can’t be physically captured properly in the cut. The resulting master is a version that purposefully doesn’t have as many volume swings and harsh highs or boomy lows. Smooth, cohesive, warm. There are also track ordering strategies which is why ballads tend to be at the end of one side and the high energy stuff up front where there is “resolution” to serve it. The mastering engineer is adjusting each song with all this in mind.
Fantastic watch thanks for sharing. I realize now how a favorite game of mine, Wario’s Woods on SNES, juices up a twist on match 3 puzzle and how dry early versions of Tetris were (succeeding despite that).
I had to take a long distance road across the US for work in the spring. I planned my hotel for the night to be around a Chipotle along the way. It’s one of the few options that isn’t synthetic and doesn’t make me feel the awful “I’m traveling for work trash diet” feeling. Some actual vegetables and protein at least. Didn’t have time to seek out local places and delivery is hit or miss. Is it possible their prices are up because the cost of “actual food” has gone up? Beyond that you’re eating low quality borderline carnival food (sugar, cheese blob, deep fried) and mono sodium glutamate loaded Christian faith based chicken sandwich offerings.
One thing I can’t find anyone mention in reviews - does inference screech to a halt when using large context windows on models? Say if you’re in the 100k range on gpt-oss. I’m not concerned about lightning inference speed overall as I understand the purpose of the spark is to be well rounded / trainer tuner. I just want to know if it becomes unusable vs reasonable slowdown at larger contexts. That’s the thing people are unpleasantly surprised to find about a Mac Studio which has prevented me from going that route.
Claude Code is still good but I don’t TRUST it. With Claude Code and Sonnet I’m expecting failure. I can get things done but there’s an administrative overhead of futzing around with markdown files, defensive commit hooks and unit tests to keep it on rails while managing the context panic. Codex CLI with gpt-5-codex high reasoning is next gen. I’m sure Sonnet 5 will match it soon. At that point I think a lot of the workflows people use in Claude Code will be obsolete and the sycophancy will disappear.
The US government might swoop in. Uncle Sam could provide a bailout on the grounds of national security if it starts to fall apart for the major players. The long play is governments investing right?
‘Together, raise and deploy a national start-up fund. With local as well as OpenAI capital, together we can seed healthy national AI ecosystems so the new infrastructure is creating new jobs, new companies, new revenue, and new communities for each country while also supporting existing public- and private-sector needs.’
https://openai.com/global-affairs/openai-for-countries/
I don't have access but it seems you can impose a friend into a video? Are we not rightsholders to our own likeness? It seems like a person should be able to block a video someone shares without their consent or earn revenue then if their likeness is used.
A really did few days ago gpt-3.5-fast is a great model for certain tasks and cost wise via the API. Lots of solutions being built on the today’s latest are for tomorrow’s legacy model — if it works just pin the version.
I work with Logic Pro X often. I bought an iPad Pro M4 and the Logic version for it is really compelling. Touch faders and the UI are well thought out. The problem is they want me to subscribe to use it. I wish I could just outright purchase it for $300.
I have optimism for (emphasis) open source LLMs and maybe it’s a counter to the things he’s concerned about now. Now we can have inexpensive device, running an open source model, offline, holding a compressed version of human knowledge (some hallucinated), that adapts to the user’s needs, is ad free, has a Wikipedia level of neutrality, no tracking of the user or blocked IPs and maybe most importantly no other people within it. Some serious contraband depending on where you are in the world and a relief maybe? The organizations that produce new ones with warped bias still will have to compete with existing models that it can be eval-ed against.
I do think they’d be better focusing Pulse on business users. Connecting information in a workplace for those with ChatGPT enterprise in this way would be welcome. If I can get notifications that span all the various information sources, infer urgency, and overnight patterns of activity of a distributed team that could have value. With zero retention policies and control over / opt in to what can be connected it’s not as invasive. I do understand the point made that it could be used to evaluate you too by someone with your analytics of time spent etc.
All these years later another thing from Zuckerberg to enable playing “hot” or “not” targeting the women around the office with his buddies.
Every owner will “Share your contacts” then do the work of labeling their friends and family in every which way for Meta. Even if those friends and family don’t want to be on social media it’ll be stored.
In the future Meta will just plan to attend a senate hearing, apologize for the misuse of that data and pay a fine.
I have had clinical OCD for decades. It becomes part of the brains default mode network (DMN) and flares up for me when frazzled, stressed or neglecting health. I’ve learned to cope, know why it happens, that it’s harmless and it’s faded with age. I recommend extreme physical exercise that burns off the associated anxious adrenaline and “reminds” the body what relaxed feels like while knocking loose stuck attention. You simply can’t focus on much else when your legs are burning huffing and puffing. There are great benefits of activating the parasympathetic nervous system, releasing natural endocannabanoids and endorphins, experiencing muscle soreness to help you feel grounded and most of all “exercising” the attention system. In time the DMN ruminations will go down exercise causes known changes to brain blood flow, activating different regions, and promoting nueroplasticity/rewiring is part of that. Get sleep to keep from getting frazzled and eat healthfully (actual nutrition, hire someone). Human-generated content may contain personal bias O_O
It takes time. There are cycles of “Oh wow!” “Oh wait...” “What if?” and “Aha!” Each of those has made me more effective and resulted in reliable benefits with less zig zagging back and forth.
For anyone that’s applied one of these at what level of autonomy are you using it? And in what setting? Greenfield?
I see one mention brownfield development. Has anyone with experience using these frameworks fired up Claude Code on enterprise software and had confident results? I have unchecked access to Claude Code at work and based on personal agentic coding I’m sure they do aid it. I have decent but not consistent results with my own “system” in our code base. At least until the front end UI components are involved even with Playwright. But I’m curious — how much litter is left behind? How is your coworker tolerance? How large are your pull requests? What is your inference cost? How do these manage parallel?
The README documentation for many have a mix of fevered infomercial, system specific jargon, emoji splatter and someone’s dad’s very specific toolbox organization approach only he understands. Some feel like they’re setting the stage to sell something…trademarked!? Won’t Anthropic and others just incorporate the best of the bunch into their CLI tools in time?
Outside of work I’ve regularly used a reasoning model to produce a ten page spec, wired my project with strictest lint, type check, formatter, hooks, instruct it to check off as it goes and do red green TDD. I can tell gpt-5 in Cursor to “go”, occasionally nudge to stay on task and “ok next” then I’ll end up with what I wanted in time plus gold plating. The last one was a CLI tool for my agent to invoke and track their own work. Anyone with the same tools can just roll their own.
Now they can demand that Intel investigate internal “waste fraud and abuse” since it’s partially owned by the government. It goes from “the government should run more like a business” to the government telling businesses “run more like the government running more like a business”
There are schemes happening in interviews too. I've been doing many technical interviews for a senior role with off shore candidates lately. They come to me after passing a challenging online test that has controls to check for browser focus lost, opening new tabs etc. I think it may require a camera to be on too. Even with that more than half of the candidates can't code.
Our interview usually starts with them breathlessly reading from a script out of the corner of their eye. I'm ok with notes to make sure you hit some high points about yourself even in person. Nervousness shouldn't disqualify a talented person. But with the coding part I've gotten exasperated and started asking these senior candidates to share their screen and do a fizz buzz exercise live in a text editor in the first few minutes. If they struggle I politely end the interview on the 15.
One candidate cheated and it was interesting to watch. In the time I sent the message in Zoom and them sharing their screen, just a few seconds, they had either queried or LLM-ed it on their phone or another computer, had someone off screen or in the same room listening and sharing the answer on another monitor or something else. Whatever it was they turned their head slightly to the side, squinted a bit and typed the answer in Java. A few syncopated characters at a time. When asked what modulo was they didn't know and couldn't make any changes to it. It was wacky. In retrospect I think it was them reading the question out loud to an LLM.
I'm waiting for the candidate who has someone behind them with the same shirt on pretending to be their arms.
The way they will incorporate an absurd mix of expressive poetic technical and satirical writing in the same piece — to the point of belaboring it and wearing you down until you can’t help but laugh is what I love. Compendiums off Amazon used books are about $8 I bought a stack a few years ago. “What Makes Anna so Beautiful in the Moonlight?” is a favorite for some reason (nerd explains beauty). Also the Onion Film Standard “The Onion Looks Back at E.T.” Maybe this means Nathan Fielder will resurrect his short lived hardcopy newspaper “The Diarrhea Times” too if there’s an appetite!
I noticed the example of ln command symbolic linking of the files on the front page. I think it’s interesting how the cli aspect of agent coding and tooling is pushing command prompt / powershell users to become familiar with the *nix shells and commands. In the enterprise Microsoft Visual Studio IDE world some of its very alien. I regularly do tutorials for team members that just think of the cli as a place where they execute ps1s, haven’t used wsl2 or git bash profiles or have limited exposure by way of dealing with containers. Not a criticism.
Same and if I need to share a bit often times I just highlight it and tell an LLM to blast out a markdown table, render preview and screenshot it to someone.
Yes o3 was an inflection point. Both modes of 5 are performing poorly on the IQ test compared to o3 https://www.trackingai.org/home That test best reflects my experience and results from practical use cases with the reasoning models when planning specs, bug finding, ideation, and deep research. It’s great at tool use as well in scripts. What I like least about the release is no transparency about the “routing” taking place. Give me all the options on the system card to pick from https://openai.com/index/gpt-5-system-card/ and I don’t want to have to start telling it “ultrathink” or other magic words to affect routing. To be fair though I haven’t tried 5 in reasoning mode beyond Cursor. But now I see o3 is only part of the Pro plan. If 5 reasoning is supposed to be better why would o3 and o3-pro still be a specialized models for Pro customers? I’d like to see some side by side prompts I might go back and test that.