Native here: you’re correct it’s a weird sentence and turn of phrase in English.
That said, can/cannot is a flexible word in English and we could take it to mean “Anyone discussing the significance of JUMPSEAT [accurately] [should never] understate it.”
But I think he meant overstate in this case. Or maybe he hated JUMPSEAT, thought it sucked and put that right out there in the press release.
They do build their own software, though. They have a large body of stuff they make. My guess is that it’s done to stay current, inform design and performance, and to have something to sell enterprises along with the hardware; they have purposely not gone after large consumer markets with their model offerings as far as I can tell.
I mean, they tried. They just tried and failed. It may work out for them, though — two years ago it looked like lift-off was likely, or at least possible, so having a frontier model was existential. Today it looks like you might be able to save many billions by being a fast follower. I wouldn’t be surprised if the lift-off narrative comes back around though; we still have maybe a decade until we really understand the best business model for LLMs and their siblings.
I think you are right. Their generative AI was clearly underwhelming. They have been losing many staff from their AI team.
I’m not sure it matters though. They just had a stonking quarter. iPhone sales are surging ahead. Their customers clearly don’t care about AI or Siri’s lacklustre performance.
> Their customers clearly don’t care about AI or Siri’s lacklustre performance.
I would rather say their products didn’t just loose in value for not getting an improvement there. Everyone agrees that Siri sucks, but I’m pretty sure they tried to replace it with a natural language version built from the ground up, and realised it just didn’t work out yet: yes, they have a bad, but at least kinda-working voice assistant with lots of integrations into other apps. But replacing that with something that promises to do stuff and then does nothing, takes long to respond, and has less integrations due to the lack of keywords would have been a bad idea if the technology wasn’t there yet.
We do know that they made a number of promises on AI[1] and then had to roll them back because the results were so poor[2]. They then went on to fire the person responsible for this division[3].
That doesn't sound like a financial decision to me.
I think of football as one of the killer apps for tv. Baseball is one for radio: baseball is almost always better over radio; there’s lots of space to do something else at the pace of most baseball games.
The most fun I’ve had with regards to baseball is either getting tanked watching the local team in clllege on cheap beer and a short walk to the stadium, or a team-management game for Wii where the main focus was building your team and playing the whole 200+ game season. There was still a game of sorts for the actual play, but it was by no means the focus in terms of presentation nor mechanics.
My grandmother (b. 1906)—a very intelligent lady (held her J.D.)—always had MLB game the broadcast tuned in anytime I was visiting her. The tempo and information needs of the game makes it perfect for delivery by radio.
Baseball is killer in-person but it's also pretty nice to just have on TV. There is nothing else like the tension of a critical at-bat.
Football is actually really really weird for a spectator sport and, I think, is generally presented very poorly. 80% of the game is deciphering opposing formations to determine what they each are predicting the opposing formation is about to try to do.
Football is just a really complicated sport; one reason I think it’s popular is that it’s fun on a visceral physical individual level (“Wow, look at that run/block/hit/kick”) for ‘beginners’ watching and is also intellectually engaging as you learn more, “wait, how did they shift the secondary just now??”
Baseball - I like it in lots of forms, too. But I think a good radio announcer can get you most of the fun out of a critical at bat narrating.
I recently watched Brockmire with Hank Azaria, and thought it was funny how the actual game announcing were just mere interruptions to whatever else he was talking about at the time. Which is pretty much how I find watching sports at a bar when there's multiple screens with multiple games and people you're with not actually into any of it.
Football is so unique in that the way it’s presented makes it almost impossible to understand what’s going on. There are a million rules, which even die-hard fans don’t understand. And the broadcast doesn’t even make an attempt to explain or even show the offensive or defensive formations and plays being chosen.
It feels like what we’re shown on tv is a very narrow slice of what’s going on. We see the ball moving down the field but have no idea what the coach or quarterback is doing. Somehow it’s still an incredible watch though.
I went to a ball game to watch a buddy's kid throw a first pitch. I didn't know that there was more than one first pitch. We sat around talking after their first pitch and it was already the 3rd inning before I realized the game had actually started. I must have missed the "play ball" announcement.
Every MLB game I've ever been to has had that happen lol. 3 'first pitches' and then the actual start of the game happens with absolutely zero fanfair, so it's very easy to miss. In general they announce very little audibly about the actual game, it's a very different experience from watching on TV.
The first time I saw baseball from box seats after 30 years of upper deck, it was two different games. While you couldn’t hear, seeing the players speak to one another was a blast.
Cricket is another where the radio broadcast was always better, as you could have the TV on in the background but listen to the radio, only looking up when something happens.
The commentators, particularly for the Aus/England Ashes series were always better with the likes of Agnew and the now retired Blofeld providing much better commentary.
Lots of Olympic events seem well suited to the format. BMX racing, freestyle skiing, luge/skeleton, and a variety of track and field events all have runs that last for less than a couple of minutes. Not sure if there’s anything comparable in the realm of professional sports besides highlights
I remember that Turkish guy and the Korean woman who looked like total badasses in the shooting competition in the Olympics, the Turkish guy because he wasn’t using all the fancy gear, and the woman because she looked like someone out of a James Bond movie. Those highlights were just too cool.
That's actually an interesting question. Table tennis, maybe? Each volley is the right length for a TikTok video, and some of them (certainly not all) have spectacular long-distance lob+smash plays.
Seems like it plays well with vertical video orientation too.
post-game Chess analysis, diving, skis/snowboards/bikes/skateboards/etc doing tricks, any sports bloopers or amazing single plays (like the dad holding a beer and a baby who manages to also catch a foul ball, racing crashes, curling throws...
that may be. but that's like saying, "XYZ is a killer app for vinyl" haha.
football as a televised spectator sport? trending down. it's not dead, but where growth is measured, it is not good. the cultural thing this guy is talking about in the article, it's going away. fewer and fewer people every year value the aesthetic experience he is describing.
TV ownership? trending down. they've never been cheaper for a reason. trend for TV production since peak TV? down.
football as a gambling product? up. okay, do you see what i mean by bad growth? football mediated as betting stats on apps? up. draftkings, polymarket, ESPN fantasy app ARPPU? up. ESPN streaming app ARPU? down. comcast? hated, down, everyone is cheering for it to go down. do you see?
there is no way to talk about specific instances of football (and stadium sports') cultural weaknesses without sounding really cringe. maybe just, "who cares?"
I get that you don't like football and you don't like television, which is fine.
As someone who's been analyzing video content industry trends for a few decades now, I just want to let you know you've reached some incomplete or misleading conclusions based a variety of category errors and assumptions. Traditional living room televisions are just one way of consuming video content. And "over-the-air broadcast" is just one way of distributing video content. Assuming broadcast television viewership shrinking also means less video is being created and consumed is like assuming music consumption is down because CD sales are down or the printed word is dying because fax machine sales are down.
The reality is quite the opposite. Video content creation, distribution and consumption are all growing at very high rates and have been for a long time. The industry puts a lot of effort into reproducible, audited measurement and has developed deep understanding of how viewership has shifted and multiplied across video consumption platforms, consumption modes, and distribution channels - ranging from streaming long-form to social snacking. While it's true that broadcast television is shrinking and traditional living room TV sales are down, far more video content is being created, distributed and consumed today than ever before, and not by a little - the growth trends are explosive regardless of how we count: viewers, views, hours, titles, revenue or reach. All the metrics measured across the entire video content lifecycle reflect the same incredible growth.
I suggest you focus on the myriad ways video content can be bad, is getting worse or has negative effects on kids, culture or human progress. But arguing video isn't growing is neither accurate nor necessary to support your point.
Do you have sales or survey data to support this claim? I’m willing to believe individual households might be less likely to purchase TVs, but my understanding is that manufacturers are producing as many or even more screens than ever, though that might be for commercial or business use. Incidentally, it’s efficiency from this scale that allows manufacturers to sell televisions at such low prices, not a lack of demand.
haha well you wouldn't have thought that at some point in the past, tv set ownership was like 100%, so of course it has gone down, but it's still trending down. this also seems kind of obvious to me.
Hmmm. NFL revenue continues to grow, with something like half of it from media licensing. So, I think you’re wrong. My son watches NFL on his laptop, but there’s little to distinguish how he watches it from the experience he would have on a Samsung tv - he’s not like in some chat group trying to get Ochocinco to write his name on a jersey - he’s passively consuming an edited video feed of a football game with commentary.
i didn't say the NFL made less money, because i'm not stupid. i'm trying to describe a secular trend so i'm comparing the revenue growth in different media companies. looking at this table, a simple way to interpret this is, kind of obviously, netflix isn't really about presenting on TVs per se, they make a media platform, which performed way better than the NFL did, almost 10:1, which is really reinforcing my point no? for every 1 dollar someone gives NFL, consumers give Netflix 9. see? to me that is a trend going down, even if to you, it is a trend going up. depends what your benchmark is!
another POV: other people do a better job at making NFL content than the NFL does, which is what you are saying your son is consuming. and listen, honestly, ask him if he or his friends bet on football...
And I would be VERY interested to see if that increasing NFL revenue was due to people spending more hours watching/attending or because the NFL have a ton of deals with all the online gambling sites whose revenue is going apeshit.
My gut feel is that people are actually spending a lot fewer hours watching NFL games.
Even when I have been at parties where an NFL game was on, the younger crowd were diddling with their phones--generally playing games. No judgment, it's not like phone games are inferior to vegetating in front of the boob tube--but I suspect that while an NFL game may be "on", it's not really the "focus" anymore.
> football as a televised spectator sport? trending down.
AFAIC, NFL football is almost always the top 99 out of 100 most viewed television programs in the US every year. The Oscars usually manage to claw onto the list and in election years a couple presidential debates make it, displacing a few regular season games. Since your claim conflicts with my current understanding, I just had AI do a quick search of recent credible sources. Here's the summary:
> "The claim that football (American football/NFL) as a televised spectator sport is trending down is not true based on recent data.
>Regular season NFL viewership saw a minor dip of about 2.2% in 2024 (averaging 17.5 million per game), but rebounded strongly in 2025 with significant gains—averaging around 18.7 million viewers per game (up 10% from the prior season in some reports, marking the highest in 36 years or second-highest on record when including updated measurement methods like Nielsen's Big Data + out-of-home viewing). Networks like CBS, Fox, NBC, and Amazon all reported year-over-year increases, with streaming platforms showing particularly strong growth.
> Super Bowl audiences continue setting records: Super Bowl LIX (2025) averaged 127.7 million viewers (up 3% from the previous year), marking consecutive record highs. Playoff games, including wild-card and divisional rounds, also showed double-digit increases in multiple cases. While some earlier seasons had slight declines (often tied to factors like election years or measurement changes), the overall trend since 2024-2025 has been upward, reinforcing the NFL's position as the dominant U.S. televised sport."
Your impression may arise from shifting measurement platform data due to increasing out-of-home, mobile, streaming, DVR, etc viewership. Just comparing traditional old-school Nielsen in-home diary data alone hasn't been accurate for over a decade. Even if we discount recent cross-platform measurement data, the overwhelming dominance of NFL football is also well supported by the audited financial reports of what broadcasters and streamers pay the NFL and further by what advertisers pay for slots. The sheer money being paid dwarfs all other sports and types of television programming (news, drama, comedy, etc). The recent dramatic growth of legalized sports gambling in the US will likely push NFL viewership across all platforms and formats even higher.
> all TV broadcast is growing like 1/10th the rate as Netflix did in the past decade.
Okay, but that undermines your earlier point. The NFL isn't tied to or limited by 'broadcast television'. NFL football is simply 'video content', but not just any video content, it's the hottest video content of all time - no matter how its distributed. Streaming is now the fastest growing distribution channel for video content, so it's also the fastest growing channel for live football video content. Netflix is paying big bucks to stream some live NFL games - with plans to increase next year. And Amazon Prime is already a major 'network' for NFL with Thursday Night Football. Industry analysts report NFL football is by far the single most expensive content/hr for Netflix and Prime and is a major loss leader for both. They're paying the NFL far more than the broadcast rights are worth as a way to 'buy' more of the subscriber growth you find so impressive. Netflix (and Amazon Prime) aren't 'beating' NFL football, they've surrendered and are joining them (at a loss).
Disney Plus tried to bid on NFL streaming rights but NFL is so expensive it's a hugely risky way for streamers to buy viewers, so Disney dropped out and recently did a deal for exclusive US live streaming rights for a much smaller sport than NFL - F1 racing. Bottom line: live sports is the biggest, most consistent driver of video content viewership - and always has been. NFL is by far the biggest video content sport - and always has been. It's been true for over 50 years, from traditional over-the-air broadcast, cable television (in the 80s NFL rights made ESPN the most valuable cable channel), satellite (in the 90s out-of-market NFL games were the largest driver of DirectTV & Dish growth) and now it's a key growth vector for streamers.
Streaming isn't a threat to the NFL, it's the NFL's biggest growth channel. In fact, the real limit on the NFL's future growth isn't distribution at all. It's already so dominant in the U.S, it has no competition close enough to be relevant. The NFL's only remaining limit is, quite literally, the size of the U.S. population. That's why the NFL's been investing huge sums trying to establish NFL football elsewhere in the world. It's their single biggest growth priority - because they're already the absolute, undisputed king of broadcast, cable, satellite and streaming in the U.S.
i didn't even say that nfl viewership is down. i think most new viewers are disproportionately gamblers, and children that don't get to choose what to watch haha.
Tao’s doing a lot of related work in mathematics, so I can say that first of all literature search is a clearly valuable function frontier models offer.
Past that, A frontier LLM can do a lot of critiquing, a good amount of experiment design, a check on statistical significance/power claims, kibitz on methodology..likely suggest experiments to verify or disprove. These all seem pretty useful functions to provide to a group of scientists to me.
You can read about this basically everywhere - the term of art is agent orchestration. Gas town, Claude’s secret swarm mode, or people who like to use phrases like “Wiggum loop” will get you there.
If you’re really lazy - the quick summary is that you can benefit from the sweet spot of context length and reduce instruction overload while getting some parallelism benefits from farming tasks out to LLMs with different instructions. The way this is generally implemented today is through tool calling, although Claude also has a skills interface it has been trained against.
So the idea would be for software development, why not have a project/product manager spin out tasks to a bunch of agents that are primed to be good at different things? E.g. an architect, a designer, and so on. Then you just need something that can rectify GitHub PRs and bob’s your uncle.
Gas town takes a different approach and parallelizes on coding tasks of any sort at the base layer, and uses the orchestration infrastructure to keep those coders working constantly, optimizing for minimal human input.
I'm not sure whether there are parts of this done for claude but those other ones are layers on top of the usual LLMs we see. This seems to be a bit different, in that there's a different model trained specifically for splitting up and managing the workload.
Upshot: Steve thinks he’s built a quality task tracker/work system (beads), and is iterating on architectures, and has gotten convinced an architecture-builder is going to make sense.
Meanwhile, work output is going to improve independently. The bet is that leverage on the top side is going to be the key factor.
To co-believe this with Steve, you have to believe that workers can self-stabilize (e.g. with something like the Wiggum loop you can get some actual quality out of them, unsupervised by a human), and that their coordinators can self stabilize.
If you believe those to be true, then you’re going to be eyeing 100-1000x productivity just because you get to multiply 10 coordinators by 10 workers.
I’ll say that I’m generally bought in to this math. Anecdotally I currently (last 2 months) spend about half my coding agent time asking for easy in-roads to what’s been done; a year ago, I spent 10% specifying and 90% complaining about bugs.
Example, I just pulled up an old project, and asked for a status report — I got a status report based on existing beads. I asked it to verify, and the computer ran the program and reported a fairly high quality status report. I then asked it to read the output (a PDF), and it read the PDF, noticed my main complaints, and issued 20 or so beads to get things in the right shape. I had no real complaints about the response or workplan.
I haven’t said “go” yet, but I presume when I do, I’m going to be basically checking work, and encouraging that work checking I’m doing to get automated as well.
There’s a sort of not-obvious thing that happens as we move from 0.5 9s to say 3 9s in terms of effectiveness — we’re going to go from constant intervention needed at one order of magnitude of work to constant intervention needed at 2.5x that order of magnitude of work — it’s a little hard to believe unless you’ve really poked around — but I think it’s coming pretty soon, as does Steve.
Who, nota bene, to be clear, is working at a pace that he is turning down 20 VCs a week, selling memecoin earnings in the hundreds of thousands of dollars and randomly ‘napping’ in the middle of the day. Stay rested Steve, keep on this side of the manic curve please, we need you.. I’d say it’s a good sign he didn’t buy any GAS token himself.
> Stay rested Steve, keep on this side of the manic curve please, we need you
This is my biggest takeaway. He may or may not be on to something really big, but regardless, it's advancing the conversation and we're all learning from it. He is clearly kicking ass at something.
I would definitely prefer to see this be a well paced marathon rather than a series of trips and falls. It needs time to play out.
Yep, it works. Like anything getting the most out of these tools is its own (human) skill.
With that in mind, a couple of comments - think of the coding agents as personalities with blind spots. A code review by all of them and a synthesis step is a good idea. In fact currently popular is the “rule of 5” which suggests you need the LLM to review five times, and to vary the level of review, e.g. bugs, architecture, structure, etc. Anecdotally, I find this is extremely effective.
Right now, Claude is in my opinion the best coding agent out there. With Claude code, the best harnesses are starting to automate the review / PR process a bit, but the hand holding around bugs is real.
I also really like Yegge’s beads for LLMs keeping state and track of what they’re doing — upshot, I suggest you install beads, load Claude, run ‘!bd prime’ and say “Give me a full, thorough code review for all sorts of bugs, architecture, incorrect tests, specification, usability, code bugs, plus anything else you see, and write out beads based on your findings.” Then you could have Claude (or codex) work through them. But you’ll probably find a fresh eye will save time, e.g. give Claude a try for a day.
Your ‘duplicated code’ complaint is likely an artifact of how codex interacts with your codebase - codex in particular likes to load smaller chunks of code in to do work, and sometimes it can get too little context. You can always just cat the relevant files right into the context, which can be helpful.
Finally, iOS is a tough target — I’d expect a few more bumps. The vast bulk of iOS apps are not up on GitHub, so there’s less facility in the coding models.
And any front end work doesn’t really have good native visual harnesses set up, (although Claude has the Claude chrome extension for web UIs). So there’s going to be more back and forth.
Anyway - if you’re a career engineer, I’d tell you - learn this stuff. It’s going to be how you work in very short order. If you’re a hobbyist, have a good time and do whatever you want.
I still don't get what beads needs a daemon for, or a db. After a while of using 'bd --no-daemon --no-db' I was sick of it and switched to beans and my agents seem to be able to make use of it much better, on the one hand its directly editable by them as its just markdown, on the other hand the CLI still gives them structure and makes the thing queryable
Steve runs beads across like 100 coding environments simultaneously. So, you need some sort of coordination, whether that's your db or a daemon. Realistically with 100 simultaneous connections, I would probably reach for both myself. I haven't tried beans, thanks for the reference.
yeah that does make sense that these choices are related to it being a big part of gastown, still I feel it would be much more sensible to make a different abstraction separating beads core features from the coordination layer
That said, can/cannot is a flexible word in English and we could take it to mean “Anyone discussing the significance of JUMPSEAT [accurately] [should never] understate it.”
But I think he meant overstate in this case. Or maybe he hated JUMPSEAT, thought it sucked and put that right out there in the press release.
reply