#Unforced Five

[laz-kay]

One of the greatest developments in Bob’s Burgers imo is that Rudy is just… there now. There’s this unspoken rule around Rudy that he’s practically an honorary Belcher, and he casually slots himself into their lives so effortlessly. His character has been so well developed over the years that he now fits into any circumstance the Belcher’s find themselves in.

He ditched two separate Thanksgiving meals with his Mom and Dad just to see how The Belcher’s Thanksgiving played out, ended up eating with them, and probably stayed the whole evening.

He’s made multiple appearances in the kid’s fantasies, especially Louise’s. Gene and Tina even included Rudy in their renditions of Louise’s future, because of course Rudy is going to be there. Why wouldn’t he be?

When his family’s dinner took a turn for the worst, his first port of call was the restaurant. He knew he’d be safe there and would automatically click into whatever situation the Belchers had going on. He’s such a natural fit.

Tina, Gene and Louise have always sat by themselves at lunch; it's their tradition to hang out over their breaktime. Rudy sliding his way into their conversation here feels so natural and unforced, like he was always a part of their family. They're so used to him being around, they don't even question when he turns up out of nowhere.

The only other character who comes close to what Rudy has with the Belchers is Teddy. Teddy has been a constant since season 1, but he is yet to have an episode based solely around him without the Belchers looming over the storyline. "Thanks-Hoarding" was about Teddy, sure, but the Belcher's still had a huge role to play within that episode and was from their POV as opposed to Teddy's.

"The Amazing Rudy" was from his POV throughout, and the Belchers were a secondary narrative. He's the first character outside of the main five to have an episode of their own, and I think that's a statement in itself.

Idk, man. I just really think he's neat and deserves the absolute world. I could talk about him all fricken day tbh.

[jtl-fics]

Fluent Freshman - Part 05

PREVIOUS

It might be objectively funnier if no one knows Fluent Freshman’s first name (even me because there’s some funny ideas but nothing as funny as “UNKNOWN” Smith) aside from Wymack.

Even the announcers just refer to him by his last name (maybe his first name is very foreign and hard to pronounce, maybe that’s why he got into foreign languages, maybe it’s so bland that the announcers just can’t his full name without yawning, maybe it’s just that someone else on the Foxes has the same first name so there’d be confusion, who knows) This only occasionally creates confusion because the Terrapins and the Ravens both have someone with the same last name as him.

Either way after one-two hit confidence crushing combo of the CVS girl knowing he has tummy trouble and the conversation with Neil where Neil asks about his stomach ulcers FF utilizes all of his years of learned behavior to fully fade into the background. Wallflower? No man I’m Wallpaper you cannot see me.” This has unintended consequences.

1. He has become undetectable which means that he is now, once again, hearing all sorts of relationship talk but now it’s not just between Captain Neil and Andrew Minyard and their surprisingly soft private life.

He is subjected to Nicky gushing to Erik in German. He has to listen to Kevin talk to someone named Jean in French (he only finds out later that they’re NOT dating but it sounds like date talk...though most of his experience is listening to Captain Neil and Andrew Minyard loving say they hate one another so maybe he has a bad pool of knowledge.) Matt, Aaron, Jack, and some of his other freshman players all speak in English but CHRIST he wishes it was in any language he doesn’t know. (If he hears Jack say ‘Yeah sure whatever Babe.’ one more time HE’S going to slap the asshole for being so dismissive to his girlfriend.)

2. He has become objectively better at being a defensive dealer because other teams just kind of don’t notice him and pass right to him. Kevin wants to study him on what about him makes him so adept at causing unforced errors in objectively better players. Fluent Freshman, having been yelled at by Kevin more than once, takes the moment Kevin’s eyes are diverted by Captain Neil saying something and just slides out of his seat into a different seat (Middle of the bus. The least noticeable section) and puts up his hoodie and pretends to sleep. (he does not. what if he drools? It’s already hard enough getting sleep in a room with two roomies. He got to campus earliest so that he could claim a top bunk so no one could see him even if heights make him nervous.)

3. Since he has become undetectable that means that Andrew Minyard and Captain Neil have a hard time detecting him. So he has not had to endure their silent judging company. He is sure that Andrew is waiting for him to slip and he’s also sure that Captain Neil thinks he’s stupid. No other freshman gets as much homework help from the Captain as he does. Still, he doesn’t have to sit in the silent knowledge that they both dislike him enough that they want him to know it. HIs ulcers are starting to get better and the girl at the CVS (He’d have to have a whole conversation to change where his prescription is filled and he’d rather die) didn’t even recognize him when he came in for a refill on his prescription or note that he was buying pepto bismol again.

HOWEVER

This means that Neil has come back with slumped shoulders and uneaten snacks the last five times he’s gone for his usual hangout. They are never planned hangouts it’s just that FF is always in the same 3 locations because there are only so many spots on campus where your back is completely secured and that weren’t so intimidating to go into (The library is terrifying. What if he sneezes???) and for the last two weeks Andrew has been unable to find FF outside of practice either.

(It has to be stated that FF is in the EXACT same spots as before but he is in stealth mode and therefore only visible if you are a mantis shrimp.)

Andrew Minyard does NOT like this. FF is THE singular tolerable teammate. Andrew likes the decompression time  he gets when him and FF sit in quiet and do their own things. FF never asks for his attention but occasionally offers him good german literature recommendations for every stage of literacy ever since Andrew had mentioned wanting to get better at the written language so he wouldn’t have to rely on Neil or Nicky when the Monsters vacationed there next summer. He doesn’t call people friend easily but FF is a friend.

So it’s incredibly irritating that for the last two weeks he’s been completely unable to grab that friend and ask what the hell is going on. The Foxes have histories and it wouldn’t be the first time a Freshman’s past caught up with them or that one of them had something set them back into old survival habits. It’d happened with Matt, it’d happened with Aaron and Andrew, it for SURE happened with Neil, Sheena’s coping mechanisms had been the bane of Andrew’s existence the year before, some of the other freshman had setbacks this year and a few players over the years have had to leave due to the very circumstances that gave them a home with the Foxes. Andrew had never really cared as long as his people were fine.

It’s just that Andrew actually kind of cares about this Freshman, he considers him as someone who could be one of his people.

Which is why Andrew will keep stalking the campus around FF’s preferred haunts until he grabs him.

Andrew catches him outside of stealth mode (He had been using a urinal that is NOT where you don’t want people to notice you) and grabs him by the scruff. “We’re taking a drive.” he says and FF just sort of goes limp and lets himself be dragged off.

He hopes that he can write his grandma a letter telling her how much her love and care over the years meant to him before Andrew kills him. He’d even take if Andrew would only let him write it in his own blood.

NEXT

Per your requests:

@i-have-three-feelings, @blep-23

[bonny-kookoo]

If Drabble are open for WDYWFM could we see her reaction to her first night at jungkook house?

- a loving fan

Jungkook doesn't seem to be very considerate of your naturally very jumpy nature- and the fact that you've not been out of the carehome for more than five years at this point. But once you look a bit closer, it becomes clear what he's actually doing.

Some might call it a 'cold shower'- he's not masking anything of what's to come, shows you who he is and what his life looks like raw and unfiltered instead of trying to sell you an idea.

He's true to his word.

In the car you both take to drive to his home, you're stiff and unsure, never daring to make eye contact with the driver at all. You rather watch the outside world pass by the window, streets unfamiliar and strangely intimidating as you watch the bustling city around you. But you don't spiral-

Because Jungkooks hand is always in contact with you in some way, the way he casually chats with his driver natural and unforced. You don't listen in on what's said- but you know you must be a topic in one way or another.

And when you finally enter his home, you're a bit overwhelmed.

Everything smells of him, everything is so.. big, and empty, and intimidating, and expensive looking. What if you knock something down with your tail? You're in no way insured for that kind of damage, so you rather pull your fluffy tail close, keeping it tight against your body.

"You can stay in the guest room here- if you want." He offers. "If you wanna sleep in my room I can put a mattress there so you can sleep there too. It's no bother." He reassures, and you just nod stupidly as he rolls your travel bag into the living room area. "You can just relax foxy, no need to be so stiff. Go roam around, I'll go shower real quick." He offers, ruffling the top of your head for a second before he vanishes into what you assume must be the bathroom.

Standing in the middle of his apartment, you're frozen.

What if he's testing you? You don't know him. You know who he has been presenting himself as up to this point, but you don't know him, down the line. He might not be physically aggressive or an asshole, but he still has to have issues with things you don't know yet.

So when he emerges from the shower with barely towel dried hair, he finds you sitting right where he'd last seen you, your head tilting straight backwards, looking at him upside-down.

He can't help but laugh.

"Oh darling." He shakes his head, before he sits down at his kitchen counter, patting his thigh. "Come here, yeah?" He asks, and you instantly get up to walk over to him- carefully sitting somewhat on his leg, not daring to put any of your weight on him-

Though he won't have it, arms properly moving you up to sit on his leg instead, chair squeaking a but under the change of weight.

"If I scold you for something, at any point of you being with me, know that it's not because I hate what you did, or hate you for what you did." He carefully explains, gently removing your tail from your hands, letting it swing back to move without restraint. "And if I do something you don't like, know that I won't dislike you or be bothered by any of it. Boundaries are guidelines, not walls to lock you up in." He tells you, moving some hair out of your face. "If you're unsure, tell me. If something isn't clear, tell me. If you're scared, or angry, or confused, or sad, tell me. I'll do the same." He offers.

"What if I don't.. like something you do?" You ask, looking down at your hands.

"Then we'll find a compromise." He shrugs easily. "Like, let's say you don't wanna sleep alone, like I said, you can sleep in my room. Or if you don't like a certain food I make, I'll just make something else for you, no biggie." The idol says, as if it's no big deal.

"Can I.." you mumble, almost not saying it out loud. "..sleep, like.. in your bed? Next to you?" You wonder, ears pinned down in submission, and he just chuckles, grinning.

"Sure thing." He agrees, before patting your back and sliding you off his leg to stand on your own again. "But first, let's eat something. I'll choose for us, okay?" He asks, and you nod-

Making him smile, because he can spot the faintest tail wagging from you as he turns to grab his phone.

You'll be fine.

He just knows it.

[mandokero-eboy]

The democrats, being the unforced errors party, were really holding up on us. But now they're majorly fucking up by very publicly snubbing the uncommitted movement at the DNC. All they were asking was to have *one* Palestinian speaker and they couldn't do that much. Come on people. What are we doing here?

I bet whoever made that decision is the same kind of idiot who thought Biden was a strong candidate right until the minute he dropped out.

This is a key issue in the election that was hurting Biden the entire time he was running and they seem to think occasionally saying the word ceasefire and otherwise avoiding the issue as much as possible will get people who, to be clear, are morally outraged about the candidate being complicit in genocide by her involvement in the Biden admin, to vote for them.

If they had time for multiple republican speakers then they had time for a five minute vetted speech by a Palestinian American. There was literally no reason not to allow this.

[recurring-polynya]

I was complaining during yesterday's episode how much I hate the conceit that the Quincy hate each other and turn on each other constantly, simply because it's stupid. These buffoons have had 1000 years to plan this, and they just make unforced error after unforced error. Then, it occurred to me that if the Gotei hadn't gone through their Kurosaki Ichigo-inspired Power of Friendship arc, they would probably also be killing each other left and right. If the Blood War had happened five years ago, you cannot tell me that Byakuya wouldn't have just killed Hisagi and later on been like "that was someone's lieutenant? Could have fooled me." Aizen would probably be trying to quadruple cross both sides. Zaraki would have taken out at least two other captains by now and no one would have come to help him when he was lying in that ditch. That being said, I'm also kind sad Gin couldn't be here, RIP to you, sir, you would have loved the Blood War.

[yourworsttotebag]

tennis au

been watching a lot of tennis so I wrote some tennis au stuff but in the form of broadcast commentary. it's fun writing in the specific cadence they use.

🎾

"Remember Wimberley at the Australian Open at the beginning of the year, when she broke her racket, getting that point deducted that cost her the set. Disappointing finish in the third round of that tournament. This is a completely different player -"

"Night and day."

"Completely different to anything we've seen before. This is a laser-like focus."

"Yeah, she lost a lot of points to unforced errors and getting in her head, losing her temper especially. But when she's focused and the forehand is working, and the serve is working, she is just untouchable."

"The speed out there -"

"Absolutely, she's one of the fastest players on the tour right now."

"Oh, textbook backhand there from Wimberley to go up five to three. With the scream and the fist pump, that's that Wimberley intensity. She is very comfortable right now, really dialed in. Cliffgate cannot keep getting pulled into these long, drawn out rallies. She needs to use her strength to put the ball away."

"We mentioned speed, Wimberley has always been right at home with the speed and pace of hard court tournaments. There's been a big jump in her play since Gale Dekarios officially joined her coaching team. You can see him there sitting with Lae’zel Gith, she's been Wimberley’s coach for the last few years. She's got Gith on fitness and Dekarios on the mental part of her game. He's a former world number one, with a lot of knowledge, great tennis intelligence. Wimberley said after her match on Monday that she trusts him a lot. This is a very solid team."

"Yeah, you know Wimberley has been very vocal about wanting a gold medal in next year's Olympics -"

"Very vocal, ha ha."

"Yes, very - very vocal. Uh, you know she was famously kept off the podium by Hallowleaf in Tokyo. And honestly, I think most people would have counted her out for gold but she's been looking very solid in the back half of this year. The runner up at the French, her second Wimbledon title after that. Looking very comfortable here at the US Open. If she keeps this up, I think she may be looking at some new hardware in the future."

[fritzes]

semifinals thoughts for every match that doesn’t involve a violent abuser (bc fuck him, I love writing these and the other three matches deserve to be talked about)

wta:

gauff v sabalenka: a rematch of the us open final! if you took coco’s quarterfinal performance and aryna’s quarterfinal performance and put them against each other, aryna would win handily. coco’s unforced errors, especially on the forehand side, are really bringing her game down. that being said, she had some shaky matches at the us open but when it really mattered, she found her best game, kept her cool, and won. aryna is known to be the exact opposite: she plays incredibly but often lets the pressure get to her in later rounds. I think the fact that she lost the us open final to coco will absolutely be in her head the entire time. since she lost from a set up, if she wins the first set here I don’t think that will be a comfort to her (in fact, it might make things worse). even if coco continues blasting unforced errors, I have more confidence in her ability to keep her mentality stable than I do in aryna’s. like I said before the quarterfinals, though, this is new territory for aryna. she’s never defended a grand slam title before, and so far it seems to be emboldening her. we could see a new side of aryna here that doesn’t get frustrated as easily, and that’s gonna be really dangerous for anyone playing her. I fully expect this to be a close, three-set match, but if one of them isn’t at the top of their game then the other one is going to take full advantage of that and win

zheng v yastremska: qinwen is the favorite, but I don’t think this is going to be as straightforward as it looks on paper. dayana is incredibly under-ranked (I’m not sure what happened, but I think she might have had some doping accusation issues?). she has been playing some insanely good tennis at this tournament, and she will absolutely challenge qinwen. the key for qinwen here is to serve well - that’s definitely one of her biggest weaknesses, but if she can do that it gives her a massive advantage because there’s really nothing else in her game that dayana can attack. this would be a first grand slam final for either of them, so whatever their unique ways of dealing with pressure are will absolutely come into play here as well

atp:

sinner v djokovic: the one we’ve all been waiting for… this match felt inevitable from the moment the draw came out. jannik has been playing the best tennis of his life in the past few months, and he absolutely carried over the momentum from last season. of course, novak is always going to be the favorite, especially at the ao. the big question for jannik here is endurance. his history in close five-setters isn’t great, and all of his matches so far have been pretty one sided. if he wants to beat novak, his body needs to hold up. if he does that and he keeps playing the way he’s been playing (preferably with a few less unforced errors), he has a real chance, especially when you consider how much time and effort novak has spent on the court. but, novak is novak, and he’s 10-0 in ao semifinals. if jannik wants to win this match, he has to be perfect

[findmeinthefallair]

Me, dishing out short metas about the Lions in the year of our lord 2023? More likely than you think.

I revisited some scenes and fondly see some common threads as to why Red selected Alfor, Keith and Lance. The official handbook (that lines up with the era of the show before the Head of Story, Tim Hedrick, was too boxed in and disallowed from following the story bible he put together) says determination and instinct on top of skill are key traits to pilot this most temperamental, most unpredictable and most unstable out of all five Lions. She's the speed demon of the team.

Alfor followed his instincts about researching quintessence to defy Zarkon twice. We know that Keith is fine with defying anyone, even Shiro at the end of the day if he has to. In my opinion, Lance's learning point was not caring what others thought of him and stop trying to do the "cool" thing. Defying that insecurity about self and thus being able to defy others in situations: wherein he would previously not be able to make that difficult decision. In early S3 he manages to be unflustered and stable to redirect Keith towards calm leadership.

So it's about making tough calls and sticking with the consequences, something shared with being in the position of Black Paladin. In fact, Red and her pilot has had to step in to lead and redirect the leader (Black) a few times. A failsafe if you will, to step in and take the reigns if needed.

She is second-in-command but she's also the wild card Lion. While by default she is unpredictable, she is the untameable one - more so than Black, I'd say - which is in a way what's most consistent in her, and is the most predictable thing about her. She'd expect any of her pilots to also have principles strong enough to defy upper command if such an act may actually be needed for the benefit of the team.

Which is interesting because..these qualities makes her not so different from her supposed polar opposite, Yellow, who is responsible for making the team morale as unshakeable as possible - "[the Yellow Paladin's] heart must be mighty" - plus civilian search-and-rescue. A consistency in principles and morals even if it is a high price to stick to them.

(While for Blue and Green, they're opposites in the sense that Blue requires spontaneity, while Green requires meticulous observation. Where they are similar is they are crucial in gathering intel. Blue's sonar ability, Green's cloaking ability)

You may need Red for offense but also defense for the team i.e. she pushes another Lion to safety twice in the show: Keith supporting Shiro in the S1 finale, Lance sparing Allura damage in S6 (which was good, since she has her Altean healing ability for herself and also others).

I really really would've wanted to see Alfor using the Rail Gun though. I wanted to see what would piss him off enough or make him scared enough to activate it. In all his screentime he never truly lost his composure (Allura is mostly like that too, she clearly gets it from him) but that one weapon requires a huge burst of unforced passionate emotion to activate: much like how a war cry works. Fitting for the Fire element she is matched to and the guardian spirit of.

Alfor is an alchemist and more cerebral which is so different from his two successors are like. His fire burned differently but he was definitely passionate.

[theresa-of-liechtenstein]

what a fucking disaster of a day

- was shit at removing supernatant fluid from around my nanoparticles so had to do it again

- was shit at doing it the second time so just gave up and put slightly diluted sample in tube furnace

- programmed tube furnace and had lunch while letting gas flow over my sample

- turned on furnace (or so i thought) and went to start another batch of nanoparticles

- could not find fucking pipette so emailed my grad student who was doing a discussion section and asked her and she was like ‘borrow one from a random grad student if they don’t help you tell me 👿’

- FIVE grad students asked me if i needed a pipette after that i think they have a group chat

- DIDNT HAVE ENOUGH FUCKING NANOPARTICLE WHAT THE FUCK I SWORE I HAD 1 mL LAST WEEK???? IT DECREASED??????

- had to Wait For My Grad Student to show me how to treat more nanoparticle

- WE FIND OUT (1) SOMEONE CLOSED THE GAS CANISTER WHEN I WASNT LOOKING AND (2) I DIDNT EVEN TURN ON THE FURNACE RIGHT

- finally properly turned on the furnace and went to treat more nanoparticle

- couldn’t measure out 6.25 mg for like. ten minutes. was doing minute changes like a fucking fool

- had to balance out the pH with NaOH and dripped in HCl instead THEYRE BOTH CLEAR AND WE WERE PANICKING BECAUSE THE PH METER WAS GOING DOWN INSTEAD OF UP

- drove home and made SEVERAL unforced errors

- and now i fucking put my shirt down on a spot where my mom used bleach and now my new fucking shirt got bleach on it

- and now i’m crying instead of doing anything constructive like the useless shit i am

in short i CANNOT do anything right and had a complete clusterfuck and waste of a day. great job me i LOVE being stupid all the time

[stateofsport211]

Wimbledon MS Q3: Meet Your Qualifiers (1/2)

Wimbledon qualifying rounds, illustration (📸 CNN)

The condensed version in a Twitter thread can be found here.

After a grueling competition spanning 2 rounds, the Wimbledon Championships' qualifying rounds in Roehampton wrapped up with 16 best-of-five matches playing simultaneously. These qualifiers might be dangerous, but before the draw is made, it is not proper if the qualifiers are yet to be met. Therefore, before freaking out "Who are they and what are they doing to the favorites" yet again, here are some summaries from the last qualification round.

They are written in the order they advance. Beware, this is probably another long read.

Section 3: Zizou Bergs [3] def. Denis Yevseyev 6-1, 6-2, 6-2

Zizou Bergs' point to 5-1* 15-30, 1st set (left) and to break for the match at 6-2 in the 3rd set (right) (📸 Wimbledon qualifications feed)

One of the most in-form seeds, third seed Zizou Bergs, advanced to the third qualifying round after defeating Omar Jasika and Pierre-Hugues Herbert, the latter in the most scenic route possible (6-3, 6-7(4), 6-4). He faced Denis Yevseyev, who stunned 28th seed Titouan Droguet 6-4, 7-6(3) in the second qualifying round, as well as defeating Matteo Martineau in three tight sets. This match turned out to be a masterclass from Z. Bergs, where he scored an almost flawless performance throughout the match.

The third seed began this match with an early break, followed by a running forehand to generate his one-point lead before serving the first-set breadstick (6-1). His dominance continued through the second set as he took it 6-2, and somehow found his way breaking for the match thanks to his forehand winner as D. Yevseyev's balance went completely off, also taking the third set 6-2 to confirm his spot in the main draw.

Section 4: Mark Lajal def. James Duckworth [4] 6-4, 6-4, 6-4

Mark Lajal's point to *3-3 40-40 (1st deuce), 1st set (📸 Wimbledon qualifications feed)

One of the notable fast-court/indoor goats, Mark Lajal, despite having a questionable hard-court form back in the first portion of the Asian Challenger swing earlier this year, the Estonian had a decent grass-court season as he made it through the second qualifying rounds of the Surbiton and Nottingham Challengers, but he could not make it to the main draw, somehow finding himself defeating Francesco Passaro and Henrique Rocha in the two qualifying rounds. In contrast, fourth seed James Duckworth had more experience as he qualified for Halle (500) despite the loss to Hubert Hurkacz in the second round.

While this match was expected to be a tough, possible servefest at best, it turned out to come down to several crucial rallies. For instance, after a preceding unforced error, M. Lajal's forehand down-the-line error secured an important deuce before the Estonian broke to 4-3, followed by a consolidation to 5-3. While M. Lajal took the first set 6-4, J. Duckworth took a medical timeout midway, but despite his best efforts to stay competitive, he fell short in the exemplified said moments. As a result, M. Lajal became the first Estonian to qualify for a Grand Slam main draw since Jurgen Zopp became the lucky loser in the Roland Garros 2018, where he was eliminated in the third round against Maximilian Marterer.

Section 7: Quentin Halys def. Beibit Zhukayev 6-3, 6-4, 6-4

Quentin Halys' point to 3-3* 30-40, 2nd set (📸 Wimbledon qualifications feed)

Quentin Halys might have had a questionable form within the past year, but he tried to salvage it in the Wimbledon qualifying rounds a year after taking a set off Jannik Sinner (in the third round) as he defeated seventh seed Gregoire Barrere 6-3, 6-4 and survived a three-setter against wild card Jay Clarke 6-3, 3-6, 6-3, and faced a decent test in Beibit Zhukayev, who stunned Jurij Rodionov 7-6(5), 7-6(7) and 17th seed Hamad Medjedovic 6-7(3), 7-5, 6-4 in the second qualifying round. Staying aggressive on return while maintaining service game consistency became important in this match, and it proved itself as the match progressed.

After taking the first set 6-3, the Frenchman tried to stay consistent as he was spotted finishing his shots down the line: a forehand one that got B. Zhukayev ran over from his forehand side to minimize the gap, followed by a timely backhand down-the-line winner before he broke due to another unforced error from the unseeded Kazakh. By then, Q. Halys broke to 4-3 and strengthened his position with a service game hold straight away, where he successfully took the second set 6-4 right after, with a similar process from his groundstrokes neutralized B. Zhukayev's usually effective first serve to take the third set 6-4, securing his spot once again in the main draw as a result.

Section 14: Elias Ymer def. Sho Shimabukuro 7-6(5), 6-3, 6-1

Elias Ymer's point to force another deuce at *0-0, 2nd set (left) and to set up his set point to 5-3* 15-40, also in the 2nd set (right) (📸 Wimbledon qualification feed)

Despite struggling to find his form back, Elias Ymer started to rebuild his form by reaching the Oeiras 4 Challenger final back on clay (l. Jaime Faria) and has tried to maintain his form since. He defeated Lukas Neumayer in straight sets in the first two qualifying rounds before fighting thick and thin to defeat Marco Trungelliti 6-7(4), 6-4, 6-2 in an entertaining encounter. E. Ymer then faced Sho Shimabukuro, who also made the cut last year and notably made it to the second round of the Surbiton Challenger (l. Leandro Riedi), as well as having a competitive first-round exit against Cameron Norrie in the Nottingham (grass) Challenger as a qualifier before stunning 14th seed Jozef Kovalik and recent Sassuolo Challenger champion Jesper de Jong in both qualifying rounds.

E. Ymer initially had a neat start to the match as he brought S. Shimabukuro to the tie-breaker, ultimately taking the first set 7-6(4) despite starting the tie-breaker in an unideal manner (with a double fault), recovering the mini-break deficit through his groundstrokes (mostly through his forehand side) before setting up the mini-break through S. Shimabukuro's forehand mishit. The Swede had not looked back since, taking the first set 7-6(5) before scoring a backhand down-the-line winner to force another deuce before breaking early to 1-0 to start the second set. Eventually, E. Ymer outhit S. Shimabukuro through his occasional slices in between his forehands, breaking for the second set 6-3 before dominating the third set that ended with a breadstick (6-1) to secure his third Wimbledon main draw qualification after 2015 & 2022.

Section 10: Alejandro Moro Canas def. Damir Dzumhur 7-6(6), 6-4, 0-1 ret.

Alejandro Moro Canas' point to create his break points in the first game of the second set, with the earlier (left) and the later (right) attempts (📸 Wimbledon qualification feed)

Alejandro Moro Canas started to prove himself to be an all-around player (outside of his usual clay and occasionally hard-court prowess), continuing his form after winning the Rome (Italy) Challenger during the clay season, followed by qualifying for the main draw of Roland Garros (d. Mattia Bellucci). For the Challenger-level grass season, he made it as far as the second qualifying round of the Ilkley Challenger (l. M. Tomas Barrios Vera) before stunning Joao Fonseca 4-6, 6-3, 7-6(10) in a topsy-turvy match, followed by 29th Matteo Gigante 7-5, 6-2 in the first two qualifying rounds. On the other hand, 10th seed Damir Dzumhur defeated Tristan Schoolkate and Nick Hardt to make it to the third qualifying round despite his first-round loss to Joris de Loore back in Ilkley.

However, the likely physical first two sets already took a toll on the 10th seed throughout the match. A. Moro Canas then successfully took the first set 7-6(5) despite being comebacked when he should have served for the first set at 5-3, ultimately mini-breaking for the set thanks to his timely forehand winner. A working volley to Dzumhur's drop shot also secured A. Moro Canas' earlier break point to start the second set, but even though it was foiled, the Spaniard created another one through a forehand winner, which was successfully converted, and he had not looked back since. There was several close-call-related chaos midway, but the physicality of this game prompted Dzumhur to retire despite holding the first game to start the third set (1-0), sending A. Moro Canas to another Grand Slam main draw appearance.

Section 6: Lucas Pouille def. Luca van Assche [6] 7-6(4), 7-5, 6-2

Lucas Pouille's point to *5-2 40-40 (break point save) in the third set as he served for the match (📸 Wimbledon qualifications feed)

Despite some questions surrounding his comeback and inability to catch a break, Lucas Pouille got himself together in this year's Wimbledon qualification rounds and convincingly defeated British sensation Jack Pinnington Jones 6-0, 6-3 in the first qualifying rounds before stunning Jiri Vesely 7-6(6), 6-2 in a classic encounter. The third qualifying round became an all-French generational battle as he faced sixth seed Luca van Assche, who defeated both Jules Marie and Li Tu in three sets, with the third set being brought into a (match) tie-breaker.

This match turned out to test their range consistency, which became evident toward the end of the match.

Section 12: Cristian Garin [12] def. Timofey Skatov 6-2, 7-5, 6-3

Cristian Garin's point to break 4-2 in the 1st set (📸 Wimbledon qualifications feed)

Having to miss last year's Wimbledon due to his injuries, 12th seed and 2022 quarterfinalist Cristian Garin was determined to be back on track despite his questionable form throughout the past few years. Here, he defeated Enzo Couacaud 7-6(5), 7-5 in the first qualifying round before defeating M. Tomas Barrios Vera 2-6, 6-2, 6-0 in an all-Chilean battle, facing Timofey Skatov in the final qualifying round as the latter defeated Andrea Pellegrino 6-3, 6-3 and stunned 26th seed Lukas Klein 7-6(6), 7-6(5) in the first two qualifying rounds.

Somehow, although this match was thought to be slightly more competitive, C. Garin turned out to dominate from his forehand side as he capitalized on T. Skatov's previous second-serve errors, utilizing his cross-court forehand to break 4-2 before he held his serves to 5-2. The Chilean ultimately broke for the first set 6-2, and even though T. Skatov tried to raise his level in the second set, C. Garin cleaned up his act to finally take the second set 7-5 and continued his form as he took the third set 6-3 to secure his main draw qualification round for another occasion.

Section 1: Maxime Janvier def. Giovanni Mpetshi Perricard [1] 6-7(10), 7-5, 7-6(4), 7-6(5)

Maxime Janvier's points to 6-6(6-4, first set point save) in the first set and to *4-2 40-ad to create a break point in the second set (📸 Wimbledon qualification feed)

After his Stuttgart (250) and Queen's Club (500) breakthrough, both eliminated in the first and second round as a qualifier, respectively, first seed Giovanni Mpetshi Perricard carried on his breakthrough season with straight-set victories over Ugo Blanchet and Antoine Escoffier. Somehow, he found himself facing another, more experienced Frenchman in Maxime Janvier, who defeated Yu-Hsiou Hsu 7-5, 6-3 and knocked out 30th seed Emilio Nava 7-6(1), 6-3 to secure the third qualifying round appearance.

Interestingly, this match became another rollercoaster coming down to Gio's second serves and some follow-up returns. M. Janvier tried to exert more pressure through his passes to prolong the first set even if the first seed took it 7-6(10), but the former managed to come back from 1-4 down in the second set as he brought Gio off-balance thanks to his slightly deep returns, passing the latter through his forehand (and a forehand return ace beforehand) before breaking back to 4-3 several games later. Ultimately, frequent double faults and a forehand error resulted in M. Janvier taking the second set 7-5, and he started nailing these moments for the next two sets, nailing the thinnest margins as he took the third set 7-6(4) due to Gio's failed drop shot, and generating his match point through the first seed's +1 backhand error before finally taking the whole match, securing his maiden Grand Slam-level qualification for the main draw after receiving a wild card to compete in Roland Garros' first rounds between 2018-2020.

[maddie-grove]

I know it was a natural decision due to the popularity of the books, but the Bridgerton novels are in almost every other way a particularly tricky choice if you’re going to adapt a historical romance series into a serialized TV show in the 2020s, thanks to:

The sheer number of books. There’s no way Netflix was ever going to get around to all eight of the main series. This might not seem like a big deal—no one ever relies solely on the interest of book fans when a book is adapted—but tapping into the lucrative romance market had to have been a consideration, and more of that audience could have been retained if it wasn’t clear from the jump that the adaptation would fall so short of being complete.

The existence of characters who are children in early books but adults in later books. And, really, the prominence of child characters, period. It would have been a lot simpler, timing- and labor-wise, to go with a series with fewer sassy tweens running around.

The relative sameness of the books. It’s not such an issue if you’re reading the books as they come out every year or so, or leisurely catching up with them years down the line, but I think with a show it becomes really obvious that you can easily condense the eight siblings into approximately five siblings: Anthony (type-A eldest son), a Benedict-Colin hybrid (aimless dilettante), a Daphne-Francesca hybrid (chill yet proper older daughter), an Eloise-Hyacinth hybrid (less chill nonconformist younger daughter), and a Gregory-Hyacinth hybrid (baby). Make one of them kind of an outsider who doesn’t vibe much with the Bridgertonian whimsy (probably the Daphne-Francesca) and you’ve got everyone basically. I’m saying this as a Gregory girl. Unfortunately, condensing the siblings would be an even more controversial move than just letting the series trail off.

The relative scarcity of side plots (multi-book or otherwise). The only real multi-book arc (aside from foreshadowing of future romances) of any note is the mystery of Lady Whistledown, which is solved halfway through the series. Generally side plots are resolved in one book. This is tough on a serialized TV show. Other romance series have more robust multi-book plots (Elizabeth Hoyt is good at these).

The lack of diversity. Romance novels are generally very white, and also just really normative in lots of other ways, and that was even more the case in the 2000s than now. But even by those standards, the Bridgerton novels are homogenous. Obviously this can be changed in adaptation, and it was changed in adaptation (although, from what I’ve seen people say, to limited satisfaction), but more diverse historical romance series do exist.

The rape scene in the first book/season. I almost didn’t list this, because it was a total unforced error in both media, IMO. It serves the exact same function as if they’d had an ugly verbal argument about having children. It ultimately doesn’t matter to either character or their relationship in-story, and I don’t think it’s even supposed to be titillating. I was honestly shocked when the show kept it in (and made it way worse by making the victim black and the rapist white…and also made it more…something…by having Daphne do it angrily rather than as a furtive attempt at getting pregnant). Still, many romance novels have no rape scenes. Let alone one of the most notorious female-on-male rape scenes of the genre.

Again, obviously they took what they wanted from the books and added in their own stuff, and I suppose that worked out for them just fine. It’s just interesting that there are popular romance series that would have been way more workable.

[mariacallous]

The final moments of a convoluted and chaotic 24 hours of political drama that culminated in Liz Truss’s downfall began at about 11.40am on Thursday, when Sir Graham Brady slipped into Downing Street via a back entrance.

The official No 10 narrative was that Truss had instigated the meeting with Brady, the shop steward for backbench Conservative MPs. Few believe that, and even if it was the case, the power balance was much like a bankrupt calling in the administrator as the inevitable loomed.

While the precise details of what Brady told Truss remain opaque, the message was obvious: you have lost the confidence of too many MPs, the chair of the 1922 Committee told her, and if you do not go voluntarily you will be removed.

From that moment on, events travelled at speed, with every development increasing the expectation that the UK was heading towards a fifth Conservative prime minister since 2016.

Shortly before 12.30pm, Thérèse Coffey – Truss’s health secretary, deputy prime minister and close friend – was photographed arriving at No 10. Minutes later, Jake Berry, the Tory party chair, was seen walking, grim-faced and with a phone clamped to his ear, into Downing Street.

Less than an hour later black-uniformed staff carried a lectern outside No 10, and the game was up. Dressed in a royal blue suit, the gloomy expression of recent days supplanted by an uncomfortable half-smile, Truss delivered a statement of just 200 words. She had spoken to the King. A successor would be chosen within a week. She was gone.

Even in the context of the almost absurdly accelerated timescale of Truss’s premiership, how did she and the Conservative party get here given that, just the day before, the main challenge was initially seen as coping with prime minister’s questions?

As with much of the last 45 days, the answer comes down to a small amount of bad luck which is grossly magnified thanks to a toxic combination of cloth-eared party management, ideological tunnel vision and an often astounding tendency towards sheer ineptitude.

The condensed endgame had begun at 4am on Wednesday when Suella Braverman, still then the home secretary, joined the National Crime Agency on a raid in Oxfordshire connected to illicit Channel crossings.

As she returned from the raid in the back of a ministerial car, a presumably sleep-deprived Braverman used her personal email to send a government document about immigration to a supportive Tory backbencher, accidentally copying in the aide of another MP, who informed the whips about what was a serious, if not necessarily career-ending, rule breach.

The result is well documented. Truss called in her home secretary, and after what some reports later described as a standup row, Braverman agreed to resign, albeit with a departure letter dripping with barely coded contempt for the prime minister.

One of the many paradoxes of a truly extraordinary day was that Truss could, even then, probably have survived the loss of a second top minister in five days, at least in the short term.

Yes, many MPs on the right of the Tory party assumed Braverman had been toppled by Jeremy Hunt, who replaced Kwasi Kwarteng as chancellor on Friday, but the swift installation of Grant Shapps as home secretary signalled an intent to steady the ship.

Similarly, such had been the endless news wattage of Truss’s administration that the suspension earlier on Wednesday of one of her senior aides, Jason Stein, for alleged negative briefings had passed with relatively little notice.

What proved Truss’s final undoing was, instead, a sequence of events which underlined perhaps her most serious flaw as PM: the tendency to appoint loyalists, cronies and friends to top jobs, irrespective of their apparent abilities.

The challenge facing Wendy Morton, appointed chief whip after enthusiastically backing Truss’s leadership bid, had been dealing with a Labour motion in the Commons which sought to split Tory MPs by offering them the chance of a vote to effectively outlaw fracking.

The first clear error by Morton and her team was to declare on Wednesday morning that this would be a confidence issue, with MPs who failed to back the government at risk of losing the whip.

From this point events rapidly deteriorated, passing through confusion, then chaos and into outright farce. After a string of Tories pledged to rebel nonetheless, a minister stood up in the Commons to say it was not a confidence vote. As MPs lined up to vote, many simply did not know. At least one was in tears.

At various points in the ensuing melee, Morton and her allies badgered or shouted at confused would-be rebels, or according to some reports, physically shoved them into the government lobby.

Morton and her deputy, Craig Whittaker, were widely assumed to have resigned amid the bedlam, until a seven-word No 10 statement at 9.50pm said they had not. At precisely 1.33am, another Downing Street missive said the vote was a confidence issue after all, leading some rebel MPs to publicly question whether they remained Conservative MPs or not.

And all this, it must be remembered, was for a fairly routine if procedurally knotty opposition day motion, one the government ended up winning by a significant margin.

Drinking and gossiping in Strangers��� Bar on Wednesday evening, and by WhatsApp the next day, Tory MPs were in despair at the extent of unforced errors.

One cabinet minister openly blamed Morton for causing the PM’s downfall, albeit with the full complicity of No 10. “Yesterday was appalling mismanagement,” they lamented. “It was a victory with a majority of 96. If the chief whip had not lost the plot over it we would not be in this situation. Confident leaders would have just ignored games by the opposition.”

Another MP was more blunt still: “The whips lied and misled MPs to achieve an outcome on house business. The public deserves answers.”

Hours later, Truss was gone, but there was a sense among many Conservatives that the party’s troubles are only just beginning as it embarks on what could well be an utterly brutal rapid-fire race to discover who will inherit Truss’s policy poison chalice of spooked markets, spending cuts and endless broken promises.

“There is no way the party will be able to agree on one candidate,” one MP said. “We are too far gone.”

For some Tories, worse still was to come: the news that Boris Johnson might join the race. One MP was clear: “If he came back I would immediately defect to the Labour party.” There is, it would seem, much more still to come.

[anghraine]

Elizabeth, you inspire me. A while ago, I ranted to you (and asked for your advice) about my wish to major in Literature, if you remember me. I just wanted to tell you that every time I lose my motivation, I come back to your blog and just get inspired to be like you. Btw I'm getting myself prepared by familiarising myself with some of the things that might be challenging in the long run, and I'm finding myself quite frustrated with the History of English Literature. Any advice for that?

Belatedly, thank you very much! That's a lovely thing to say.

With regard to the history of English literature, that's a very broad topic. Maybe it seems more so to me because I've always tended to focus a lot on particular subjects I find compelling, and I pay probably less attention than I should to the areas that I don't specifically study or want to study—nearly all my undergrad and master's literature courses were in early modern, eighteenth-century, or nineteenth-century British literature, because I had a lot of freedom of choice and those were what I was interested in (I managed to arrange things so that I took eighteenth-century literature five times, for instance).

Apart from Tolkien classes, I think I've taken maybe three twentieth- or twenty-first century literature classes ever (the only one I much cared for was a bell hooks seminar). So I often don't weigh in on discourse around most twentieth-century literary controversies because I genuinely don't know much about them. On the other end, I've taken late medieval English lit classes but I don't know much (or care much) about early medieval literature. I've taken enough survey courses to have a general sense of most periods, but that's about it if they don't fall within my range of interests.

So I'm probably not the best person to give advice on studying English literature broadly because I did my best to specialize as often as possible, as early as possible. The only advice I can really give about English literature as a whole is to try and take it piecemeal.

This doesn't have to be breaking it down into the traditional periods of things like medieval literature, early modern literature, etc, but I would be wary of sweeping generalizations about trends or about what English literature is or what its history looks like unless the evidence is incredibly strong and you're familiar enough to evaluate it.

The attempts to create large-scale narratives often leave out the specific details that make literature interesting, and also often leave out details that are inconvenient for those narratives, the voices of marginalized people, genres the generalizer doesn't personally prioritize as much, etc. A lot of accounts of the history of the English novel do all these, for instance.

When you're looking at literature, IMO the most important thing is to look at the specific details of the particular work you're reading, before you try and fit it into any of these grander narratives or even engage closely with those narratives. Afterwards, you can dig around for context, you can learn things about the era or the literary moment that clarify things in the text, you can look at others' interpretations and learn from them, but I think it's generally better to experience your own unforced reactions to texts as much as you can before you start looking at surrounding material.

[overbearingstruggles]

i agree with the stage bants as of late but only partly. i followed the eu tour and the while they treaded carefully and we got one doozy in hamburg from jack saying they were going to play divorce parties - the banter was pretty standard and unforced. obviously not as fun as it used to be but i never noticed it being awkward. alex even called my friend group out a few times and was pretty fucking sassy most of tour. however i will say i've been to three dates this tour and i HATE the long run of the first what - five/six? - songs with no banter or hello. it used to be three and they'd have a little chat. this tour there's less breaks for them to run their mouths. it seems they're letting emmyn take the lead and saying 'well you got enough from him let's just play a show.' still a fucking great show but i miss their joking stage presence.

Yeah dude 5 whole songs! lol I even text the groupchat about that Friday. I'm relieved to hear Alex hasn't consistently been weird as fuck but that's even more curious to me. At my shows when he went to play Therapy, up onstage by himself, he really didn't say much. No thankfulness, no fan service spiel, no backstory, nothing but getting himself set up and playing. I don't expect raunchy jokes at this point and GWP can shut the fuck up and go away forever, but I keep seeing Alex leaving Jack hanging and not coming up with redirects, just shrugging and taking a drink. It straight up didn't feel like he was front manning in the way we know Alex always has with the way Jack was taking 100% of the lead to speak, literally scanning the room for something ok to say, and then still got left hanging when telling a rowdy (but not inappropriate, ya know) nostalgic story. It keeps feeling so palpable to me. And tenfold after seeing a show where he was relaxed and as normal as I could imagine.

[bylagunabay]

Unseen Warfare

FORCE YOURSELF TO PRAY

1 𝑻𝒉𝒆𝒔𝒔𝒂𝒍𝒐𝒏𝒊𝒂𝒏𝒔 5:16-18 “𝑹𝒆𝒋𝒐𝒊𝒄𝒆 𝒂𝒍𝒘𝒂𝒚𝒔, 𝒑𝒓𝒂𝒚 𝒘𝒊𝒕𝒉𝒐𝒖𝒕 𝒄𝒆𝒂𝒔𝒊𝒏𝒈, 𝒈𝒊𝒗𝒆 𝒕𝒉𝒂𝒏𝒌𝒔 𝒊𝒏 𝒂𝒍𝒍 𝒄𝒊𝒓𝒄𝒖𝒎𝒔𝒕𝒂𝒏𝒄𝒆𝒔, 𝒇𝒐𝒓 𝒕𝒉𝒊𝒔 𝒊𝒔 𝒕𝒉𝒆 𝒘𝒊𝒍𝒍 𝒐𝒇 𝑮𝒐𝒅 𝒊𝒏 𝑪𝒉𝒓𝒊𝒔𝒕 𝑱𝒆𝒔𝒖𝒔 𝒇𝒐𝒓 𝒚𝒐𝒖.”

1 Spiritual Enemies …

“I consider no other labor as difficult as Prayer. When we are ready to Pray, our spiritual enemies interfere. They understand it is only by making it difficult for us to Pray that they can harm us. Other things will meet with success if we keep at it, but laboring at Prayer is a war that will continue until we die.” Saint Agathon of Egypt

2 Heaven Taken By Force …

“If you do not feel like praying, you have to force yourself. The Holy Fathers say that prayer with force is higher than prayer unforced. You do not want to but force yourself. The Kingdom of Heaven is taken by force (Matt. 11:12).” St. Ambrose of Optina

3 Praying In Pain …

“Five minutes of Prayer when the whole body is in pain are more precious than a whole night of Praying with bodily ease.” Saint Sophrony of Essex

[jcmarchi]

AI companies are pivoting from creating gods to building products. Good.

New Post has been published on https://thedigitalinsider.com/ai-companies-are-pivoting-from-creating-gods-to-building-products-good/

AI companies are pivoting from creating gods to building products. Good.

AI companies are collectively planning to spend a trillion dollars on hardware and data centers, but there’s been relatively little to show for it so far. This has led to a chorus of concerns that generative AI is a bubble. We won’t offer any predictions on what’s about to happen. But we think we have a solid diagnosis of how things got to this point in the first place.

In this post, we explain the mistakes that AI companies have made and how they have been trying to correct them. Then we will talk about five barriers they still have to overcome in order to make generative AI commercially successful enough to justify the investment.

When ChatGPT launched, people found a thousand unexpected uses for it. This got AI developers overexcited. They completely misunderstood the market, underestimating the huge gap between proofs of concept and reliable products. This misunderstanding led to two opposing but equally flawed approaches to commercializing LLMs. 

OpenAI and Anthropic focused on building models and not worrying about products. For example, it took 6 months for OpenAI to bother to release a ChatGPT iOS app and 8 months for an Android app!

Google and Microsoft shoved AI into everything in a panicked race, without thinking about which products would actually benefit from AI and how they should be integrated.

Both groups of companies forgot the “make something people want” mantra. The generality of LLMs allowed developers to fool themselves into thinking that they were exempt from the need to find a product-market fit, as if prompting a model to perform a task is a replacement for carefully designed products or features.

OpenAI and Anthropic’s DIY approach meant that early adopters of LLMs disproportionately tended to be bad actors, since they are more invested in figuring out how to adapt new technologies for their purposes, whereas everyday users want easy-to-use products. This has contributed to a poor public perception of the technology.

Meanwhile the AI-in-your-face approach by Microsoft and Google has led to features that are occasionally useful and more often annoying. It also led to many unforced errors due to inadequate testing like Microsoft’s early Sydney chatbot and Google’s Gemini image generator. This has also caused a backlash.

But companies are changing their ways. OpenAI seems to be transitioning from a research lab focused on a speculative future to something resembling a regular product company. If you take all the human-interest elements out of the OpenAI boardroom drama, it was fundamentally about the company’s shift from creating gods to building products. Anthropic has been picking up many of the researchers and developers at OpenAI who cared more about artificial general intelligence and felt out of place at OpenAI, although Anthropic, too, has recognized the need to build products.

Google and Microsoft are slower to learn, but our guess is that Apple will force them to change. Last year Apple was seen as a laggard on AI, but it seems clear in retrospect that the slow and thoughtful approach that Apple showcased at WWDC, its developer conference, is more likely to resonate with users. Google seems to have put more thought into integrating AI in its upcoming Pixel phones and Android than it did into interesting it in search, but the phones aren’t out yet, so let’s see. 

And then there’s Meta, whose vision is to use AI to create content and engagement on its ad-driven social media platforms. The societal implications of a world awash in AI-generated content are double-edged, but from a business perspective it makes sense.

There are five limitations of LLMs that developers need to tackle in order to make compelling AI-based consumer products. (We will discuss many of these in our upcoming online workshop on building useful and reliable AI agents on August 29.)

There are many applications where capability is not the barrier, cost is. Even in a simple chat application, cost concerns dictate how much history a bot can keep track of — processing the entire history for every response quickly gets prohibitively expensive as the conversation grows longer.

There has been rapid progress on cost — in the last 18 months, cost-for-equivalent-capability has dropped by a factor of over 100. As a result, companies are claiming that LLMs are, or will soon be, “too cheap to meter”. Well, we’ll believe it when they make the API free. 

More seriously, the reason we think cost will continue to be a concern is that in many applications, cost improvements directly translate to accuracy improvements. That’s because repeatedly retrying a task tens, thousands, or even millions of times turns out to be a good way to improve the chances of success, given the randomness of LLMs. So the cheaper the model, the more retries we can make with a given budget. We quantified this in our recent paper on agents; since then, many other papers have made similar points.

That said, it is plausible that we’ll soon get to a point where in most applications, cost optimization isn’t a serious concern.

We see capability and reliability as somewhat orthogonal. If an AI system performs a task correctly 90% of the time, we can say that it is capable of performing the task but it cannot do so reliably. The techniques that get us to 90% are unlikely to get us to 100%. 

With statistical learning based systems, perfect accuracy is intrinsically hard to achieve. If you think about the success stories of machine learning, like ad targeting or fraud detection or, more recently, weather forecasting, perfect accuracy isn’t the goal — as long as the system is better than the state of the art, it is useful. Even in medical diagnosis and other healthcare applications, we tolerate a lot of error. 

But when developers put AI in consumer products, people expect it to behave like software, which means that it needs to work deterministically. If your AI travel agent books vacations to the correct destination only 90% of the time, it won’t be successful. As we’ve written before, reliability limitations partly explain the failures of recent AI-based gadgets. 

AI developers have been slow to recognize this because as experts, we are used to conceptualizing AI as fundamentally different from traditional software. For example, the two of us are heavy users of chatbots and agents in our everyday work, and it has become almost automatic for us to work around the hallucinations and unreliability of these tools. A year ago, AI developers hoped or assumed that non-expert users would learn to adapt to AI, but it has gradually become clear that companies will have to adapt AI to user expectations instead, and make AI behave like traditional software.

Improving reliability is a research interest of our team at Princeton. For now, it’s fundamentally an open question whether it’s possible to build deterministic systems out of stochastic components (LLMs). Some companies have claimed to have solved reliability — for example, legal tech vendors have touted “hallucination-free” systems. But these claims were shown to be premature.

Historically, machine learning has often relied on sensitive data sources such browsing histories for ad targeting or medical records for health tech. In this sense, LLMs are a bit of an anomaly, since they are primarily trained on public sources such as web pages and books.

But with AI assistants, privacy concerns have come roaring back. To build useful assistants, companies have to train systems on user interactions. For example, to be good at composing emails, it would be very helpful if models were trained on emails. Companies’ privacy policies are vague about this and it is not clear to what extent this is happening. Emails, documents, screenshots, etc. are potentially much more sensitive than chat interactions. 

There is a distinct type of privacy concern relating to inference rather than training. For assistants to do useful things for us, they must have access to our personal data. For example, Microsoft announced a controversial feature that would involve taking screenshots of users’ PCs every few seconds, in order to give its CoPilot AI a memory of your activities. But there was an outcry and the company backtracked.

We caution against purely technical interpretations of privacy such as “the data never leaves the device.” Meredith Whittaker argues that on-device fraud detection normalizes always-on surveillance and that the infrastructure can be repurposed for more oppressive purposes. That said, technical innovations can definitely help.

There is a cluster of related concerns when it comes to safety and security: unintentional failures such as the biases in Gemini’s image generation; misuses of AI such as voice cloning or deepfakes; and hacks such as prompt injection that can leak users’ data or harm the user in other ways.

We think accidental failures are fixable. As for most types of misuses, our view is that there is no way to create a model that can’t be misused and so the defenses must primarily be located downstream. Of course, not everyone agrees, so companies will keep getting bad press for inevitable misuses, but they seem to have absorbed this as a cost of doing business. 

Let’s talk about the third category — hacking. From what we can tell, it is the one that companies seem to be paying the least attention to. At least theoretically, catastrophic hacks are possible, such as AI worms that spread from user to user, tricking those users’ AI assistants into doing harmful things including creating more copies of the worm. 

Although there have been plenty of proof-of-concept demonstrations and bug bounties that uncovered these vulnerabilities in deployed products, we haven’t seen this type of attack in the wild. We aren’t sure if this is because of the low adoption of AI assistants, or because the clumsy defenses that companies have pulled together have proven sufficient, or something else. Time will tell.

In many applications, the unreliability of LLMs means that there will have to be some way for the user to intervene if the bot goes off track. In a chatbot, it can be as simple as regenerating an answer or showing multiple versions and letting the user pick. But in applications where errors can be costly, such as flight booking, ensuring adequate supervision is more tricky, and the system must avoid annoying the user with too many interruptions.

The problem is even harder with natural language interfaces where the user speaks to the assistant and the assistant speaks back. This is where a lot of the potential of generative AI lies. As just one example, AI that disappeared into your glasses and spoke to you when you needed it, without even being asked — such as by detecting that you were staring at a sign in a foreign language — would be a whole different experience than what we have today. But the constrained user interface leaves very little room for incorrect or unexpected behavior.

AI boosters often claim that due to the rapid pace of improvement in AI capabilities, we should see massive societal and economic effects soon. We are skeptical of the trend extrapolation and sloppy thinking that goes into those capability forecasts. More importantly, even if AI capability does improve rapidly, developers have to solve the challenges discussed above. These are sociotechnical and not purely technical, so progress will be slow. And even if those challenges are solved, organizations need to integrate AI into existing products and workflows and train people to use it productively while avoiding its pitfalls. We should expect this to happen on a timescale of a decade or more rather than a year or two. 

Benedikt Evans has written about the importance of building single-purpose software using general-purpose language models.