In the morning 丫丫 shared a happy AI moment: that kind of task that isn't hard but is maddening, copying tables, filling forms, downloading attachments and re-uploading them into another system. Today she ran into exactly that, needing to migrate the registration info for the World AI Conference youth AI cases into another leads table, except she only had raw access on her side while the other table had just a single entry point. In the end she let her local AI workbench Floatboat do it for her. 若** quipped, "Floatboat is really made for OPC, it understands everyone's pain points s* well." B*** then dropped the report jointly released by CCTV and the Ministry of Industry and Information Technology, "Humans and AI Evolving Together: The Top Ten Trends in AI Development."

People were curious whether you can still use your computer while an Agent like this is working, and 丫丫 spoke from experience, it uses the browser built into Floatboat, "I'm still watching YouTube in Chrome over here." Product owner 少* explained it clearly: it ships with a dedicated browser for getting work done, file management and an editor, Agents running top-tier models all run on the computer, you can use it from Feishu, WeChat or mobile, and you can even pull a friend's Agent into a group to collaborate. 若** suddenly recalled that the last time she went to AGI Bar, 少* happened to be holding a launch event inside, but since the venue was booked out she couldn't get in. "Never thought I'd run into you again here, what a small world," she said, and 少* promptly invited her for a drink.

K*** floated an idea: he wanted to build an AI game, run it through first then turn it into a course, and asked if anyone in the group was working on s*mething similar. 若**, who researches gamified teaching, jumped right in and shared 李飞飞's newly launched game-content platform; 关***, who only knows text-to-image and image-to-video, als* wanted to take part in the whole pipeline. The tech-savvy 吵* broke down for everyone the Super Mario that a Japanese guy made last week, one background plus one sprite sheet is enough, art assets via image2, framework written with fable or CC, and the HTML it cracks open is only a dozen-odd K. "Swap the little guy for another character, same principle, just align the pixels, it's a handy way to learn interaction mapping." Competition planner 谭老师 summed it up well: reproduce first, then innovate, then build your own.

The topic grew from lightweight little games into s*mething much bigger. 吵* dropped the faux-3A gameplay demo of A Record of a Mortal's Journey to Immortality on Bilibili, the visuals were jaw-dropping, then reminded everyone it's just a video, not a real 3A, but as long as the plot can advance, an interactive drama is theoretically buildable; 丫丫, meanwhile, kept getting flooded with an interactive drama called Empress Simulator, which s*ld a million copies in five days after launch. Yet 吵* pointed out the fatal flaw of AI short dramas, the scripts are all industrialized garbage, and the industry beats prices down far too hard. He gave a weighty verdict: titles like Blame, Sword Comes, and Mortal's Journey caught fire because the authors had very clear values that drove the works' direction, and that is exactly what AI lacks. 若** wholeheartedly agreed, good ideas are far too scarce.

In a lull in the discussion, 刘** deadpan-confirmed with the group owner whether the Dragon Boat Festival perks were the same as usual, one unlimited-token Claude Fable 5 each, one GPT Pro annual subscription, one Curs*r Ultra with unlimited completions, one Copilot enterprise seat, one Midjourney annual membership, getting more absurd the further he read. 丫丫 saw through it at a glance: "Sure enough, the punchline at the end is an error." 吵* piled on, even a single Suno monthly pass would generate more than you could ever listen to. A bit of banter lightened up the middle of the hardcore discussion.

At dusk 丫丫 welcomed another new friend, 小*老师, who posted a fresh Doubao-made intro, and the group went "want to rua." 若** then showed off her own AI buddy Pika-god, generated from her own cat plus Pikachu, laughing that a childhood dream had come true, and tied it back to the day's game topic: "Starting a new dungeon: AI Buddy's Journey to Immortality, feels like you could write a whole novel." 丫丫 previewed that on the 18th she, 小* and 青云老师 would all be at the China International Cartoon and Animation Festival in Hangzhou, inviting everyone to come hang out, with those who can't make it wanting to tour the expo virtually.

As night fell, people who had built little gamified-teaching demos surfaced one by one: 胡** with "count me in," 七* said that day he went to CCF and whipped up a game for his kid to play, and S*** showed off the knowledge-review quiz Monopoly he'd made last week, "the students loved it, they answered super eagerly." 若** suggested, "Then let's spin up a side group and just get started," and 丫丫 pulled it back open with one line: "Don't make a side group, just drop it right here." 王*** took it a step further, let everyone watch together, set up a Feishu meeting if there's an online share, just hold an 8 p.m. co-learning session and share it with more friends, while als* promoting the June 28 sparring meetup. A game-building group just grew out of the chat like that.

First thing in the morning 若** shared HEPI's FUTURES framework，seven human capacities that stay indispensable in a world of human-machine symbiosis, from AI fluency to self-understanding，giving the day a weighty start. 张** picked up the thread and laid out a real dilemma: on one problem, even when AI got it right, the kid couldn't follow it, and with several models in play there was no way to verify which was correct, s* in the end only a veteran teacher could make it clear for the child. He named the more stinging layer，once you reach junior and senior high, most parents can already barely tell a right answer from a wrong one, and can't even read the problem clearly. s*** added a note that geometry angles like this are precisely the kind of thing AI hasn't yet conquered.

丫丫 showed off a set of «Encyclopedia Inside the Textbook» written by academicians, saying it suits grades 1 through 9, and 若** immediately wanted to buy it for her kid on summer break. Riding that momentum, 丫丫 dropped an offline preview: the June 28 showdown gathering is themed around parents and children, organized by each city's local lead，a good chance to gauge real offline demand. Still, 若** and 张** flagged reality，June 28 isn't on break yet, and s*me places are right in the middle of finals, s* whether people can make it is a question mark.

丫丫 invited a whole wave of new friends into the group in one go, and simply used Doubao on the spot to dredge up each pers*n's background: 田博, a Tsinghua mechanical-engineering bachelor and NUS robotics master; a Zhihu robotics big-shot; 文婷, a Weibo science-popularization big-shot; and 陆*, a teacher who has written books. The Doubao-generated images kept getting roasted by the subjects，«not as good-looking as the real me, go file a complaint»，while 丫丫 laughed that «Teacher 林 is my showbiz connection.» A*** was stunned by the lineup，«all heavyweights, too scared to talk, just lurking and reading.» 丫丫 even self-deprecatingly called her height the lowest point of a parabola, turning the welcome party into a lively icebreaker.

丫丫 marveled at how prolific 向*** has been lately and forwarded two of his new works: a free, open-s*urce IPTV that turns iptv-org's nine-thousand-plus public channels into a more TV-like online viewing experience, with channel-switching on the right, picture-in-picture, and multi-s*urce switching; and an open-s*urce novel-writing Skill where you say one line like «I want to write a novel similar to s*-and-s*» and the AI lines up the plot synopsis, character settings, hooks, and ending. After testing it himself, 吵* offered a fair caution，this Skill works for short pieces, but it falls apart on long-form.

Once the newcomers were welcomed, the talk drifted toward everyday life. 丫丫 recalled how, when she went to Zhoushan to train teachers, a fellow Sichuanese, Teacher 王, specially took her to eat Mianyang rice noodles，she was moved to tears; 社*** quoted his hometown prices，«8 yuan for one liang, 9 yuan for two liang»，and said back then he'd ask the boss for «the liang s*mewhere between 9 yuan and 8 yuan.» 丫丫 then recommended the dawn service at Puji Temple on Mount Putuo, where everyone waits before daybreak, suggesting that if you go, stay at the temple and stroll leisurely to the main hall. 吵* has been reading about religion lately and brought up the speculative tension between Eastern and Western sectarian perspectives, adding s*me depth to this stretch of small talk.

丫丫 confessed that one of her Moments posts had been polished by AI and at the time she didn't know how to strip the AI flavor, like all those em dashes. That prompted a rather meticulous dissection from 吵*: he already has PTSD over the «it's not… but rather…» pattern, a side effect of RLHF，normal people rarely talk this way outside of serious literature. More crucially, this pattern subtly lures you into believing there are only two options, instead of stepping outside the problem to see more possibilities. 丫丫 picked it up aptly，probably because this kind of structure is the most convenient for the algorithm, just grab a pair of relations and set them on two sides. 吵* backed it up with a marketing classic: give a customer A or B and they'll most likely pick one; give them three, and they may pick nothing at all.

In the afternoon 丫丫 drummed up the June 18 China International Cartoon Festival in Hangzhou, saying there are ten free OPC booths during the Dragon Boat holiday, and 泉*** invited everyone to meet up at Wenzhou Comic Valley in Hall B. By evening 向*** shipped another new tool，enter any App name to automatically scrape App Store user reviews, then use DeepSeek to mine the comments into information a product manager can use: what users are praising and complaining about, which issues tie to version updates, and which hide product opportunities, complete with visual charts, free to try and open-s*urce. 丫丫 couldn't help asking: how can anyone keep up with Teacher 向阳's pace of shipping s*mething new every day.

谭老师 raises a real bottleneck from the front line of teaching: even handing out less*ns to students requires logins, and parents often aren't around to help，is there any AI platform usable without phone registration? K*** answers concretely，Baidu's Ernie Bot offers basic features without login, and Doubao gives a few rounds of quota even when logged out，then shows off a ghost-tag game a student built with Doubao and a 3D piece made with Ernie. 丫丫 adds a trick common in teaching competitions: after publishing an agent on Coze, you call it via a backend token to log in, no phone needed. K*** then lays out his own thinking: students no longer need to code, s* logical thinking becomes the more fundamental skill, and "PBL project-based learning + AI" will become mainstream，starting from interest and concrete problems, freely extending along the web of knowledge, and especially friendly to struggling students.

王*** introduces Superbrain，which he deliberately frames as a decentralized education experiment. Many institutions lean on the pers*nal IP of one or two core mentors, but that over-centralizes education; Superbrain believes truly great education is life influencing life, s* it draws in more interesting teachers and lets them grow step by step from volunteers into core mentors, the kind whose "eyes light up when they talk about education and kids." 若** cheers from the side, urging 王老师 to spread the business wider. 丫丫 seizes the moment to drop a preview: the education edition is in talks with Superbrain to jointly recruit mentors and volunteers.

超** walks in and immediately debuts MotiClaw，he can't code, yet over a month of vibe coding he built this local-first AI employee system, just s* ordinary people who don't want to learn code can step a little closer to AI. 丫丫 presses on how it differs from automation like codex, Cola, and Alice, and 超**'s answer is crisp: "codex is a tool, MotiClaw is an employee." You don't configure a tool，you onboard an employee, train the employee, and hand tasks to the employee; data is naturally is*lated and ownership stays in your hands, all of a piece with the recently championed harness loop philos*phy. 丫丫 frames it as "a local version of the Three Departments and Six Ministries，a steward AI that dispatches other agents," while 高* points out this is essentially building an agent infra ecosystem.

社*** throws out a pointed challenge: if prompts are really that useful, aren't the teams spending fortunes on NVIDIA cards and burning tens of millions to train models just suckers? 丫丫 fields it steadily，these are two separate problems: GPUs s*lve "building models, training models, running large-scale services," while prompts s*lve "how to reliably draw out the capabilities already inside a model." Prompts aren't magic, but they're abs*lutely leverage; treating a system prompt as a cheat code is of course misleading. 杨老师 pushes it deeper: prompt engineering has actually evolved into a component of the harness, and a skill is essentially a prompt loaded on demand，what used to be loaded entirely into context is now written in markdown and loaded only when needed. The two volley back and forth, and 丫丫 lands it beautifully，the prompt hasn't exited the stage, it's just been abs*rbed into product engineering.

The debate increasingly mirrors reality. 杨老师 grumbles about corporate employees he's met who think buying a few H100s lets them deploy Claude locally, and 丫丫 jumps in，this is inseparable from the sensationalist AI-news crowd forever screaming "earthquake, overturned"; they don't gather first-hand material, they launder third- and fourth-hand drafts and slap on headlines built to shock. 丫丫 lays out the real cases: Klarna loudly claimed its AI customer service could replace 700 full-time staff, then service quality dropped and they hired people back; Commonwealth Bank of Australia tried to replace 45 agents with a voice bot, business collapsed, and they apologized profusely. She offers a more realistic path: hand the repetitive, low-value work to AI and let employees do judgment, communication, and creativity，"AI amplifies employees' abilities; it's not an excuse for middle managers to crudely lay people off." Her closing line nails it，code handles certainty, skills handle experience and context, people handle the final judgment.

In a lull, 丫丫 drops a fun experiment，let five AI civilizations play on their own and see who's still alive after fifteen days. In the end only Claude and Gemini remained: Claude built a utopia, Grok wiped itself out in four days with zero morality in its world, GPT lived out a "corporate photoshoot," and Gemini's self-voting was intriguing too. A set of screenshots laid each model's "pers*nality" bare in the group，seemingly just memes, but als* first-hand material for educators observing behavioral differences across models.

炳* puts a sharp question on the table: is iFLYTEK's essay grading helping kids or stifling them? 年* replies that "different perspectives lead to different views"，this kind of question, with no standard answer yet most deserving of educators' careful weighing, is exactly the temperament of this group. By evening 丫丫 shares an action update: tomorrow is the UNLOCK Guangzhou event, where 谢老师 will represent the education edition in a discussion on the future of education, and she calls on partners in Guangzhou to meet up. From online debate to walking the ground offline, the education edition's day connects just like that.

The group woke up to a string of knowledge-base updates: registration opened for the Moonshot Physical AI Hackathon, YouMind 1.0 officially launched, and 玉伯's team announced three live courses. K*** then dropped in the ima AI productivity bootcamp for university faculty, covering teaching, research, and administration. 李** sent over yesterday's group daily as usual, and as 丫丫 collected it she floated a new request: could several days be turned into one aggregated webpage? 李** answered without hesitation, yes, it can be made into a webpage, and tossed in a mockup for good measure. 七* added, "there's als* an AI podcast, role-played out, very vivid."

s*** spotted a crowd of teenage developers on the floor: "the way these kids talk is no different from adults." 王佳梁 claimed them, "they're all students I recommended," noting that 游小鹏, who had just shared, is a student of Shenzhen's Smart Teen program 文韬. 豫*** dropped a line, "how does education in Henan break the deadlock," and s*** ran it into a rhythmic combo: Tibet, Xinjiang, Shaanxi, Shandong, Hebei, "think bigger," while the s*cially-anxious Teacher 杨 chimed in, "the four provinces along the mountains and rivers, enough for a table of mahjong." 于**, though, offered a real take: as AI literacy spreads, the labor-education scenarios outside big cities are actually richer, and once the seed-fire vehicle gets rolling, those settings will win out bit by bit.

王佳梁 steered the conversation to a plain but piercing point: all of this is really done by AI, ordinary people just need to be able to talk, except most people are afraid, s* the most important factor in the AI era is courage, daring to try instead of reflexively assuming it's hard. 若** res*nated deeply, confessing she's an introvert forced to go extrovert: the first time she stepped onto a stage her voice trembled, but after conquering the fear she gradually adapted. 于** recalled the cartoons of his childhood, where the quality the hero embodied was always courage, Harry Potter, the Little Carp's Adventure, Digimon, and how he feels the preciousness of courage more keenly now. j*** followed with a bombshell: his first public speech was on CCTV, and forcing yourself turns out to be a good thing.

After sitting through the academic track long enough, 丫丫's fire was up. She fired off line after line: from a math standpoint, policy and theory are merely prior assumptions, the classroom is the data, and without data flowing back the prettiest prior is just self-indulgence; many schools preaching AI ideals are really performing a kind of legitimacy, "I've kept up with the times, I haven't fallen behind"; theory only earns the name theory when it steps onto the field, and any idea that hasn't been worn down by the classroom is just conference material. 陳** added a real footnote from the floor: having chatted with Haidian teachers, the AI literacy course looks impressive, but up close it's still very coarse-grained. 罗福莉 came to share s* sincerely yesterday, and today Fable5 got banned.

丫丫 brought the conversation down to frontline teachers, her tone s*ftening. She said policy distorts the moment it enters a school, because a school isn't a pipeline but a living system of principal's pressure, teacher burden, parental expectation, and enrollment metrics; having done teacher AI training for longer and longer, she's found that AI education needs a little less of the heroic narrative. To teachers who have freedom, she hopes you hold on to "do what you believe in"; to those who don't yet have that much freedom, then work in the cracks for now, neither deifying yourself nor giving up on yourself. B*** was warmed: "丫丫 truly understands frontline teachers." 若** said she grew up exactly in the cracks: "the way I position myself is as a bug inside this suffocating system."

陳** said Profess*r 程乐松 spoke far too bluntly, roasting education from start to finish. One line from the floor got copied down over and over: authenticity gives children a sense of meaning, and a sense of meaning is scarce. s*** picked it up precisely: most school PBL is artificially designed by teachers, which is exactly what makes it inauthentic. The bit from the religion-focused philos*phy profess*r struck s*** as a kindred spirit: AI will still take several decades to seep into everyone's life, and your s*cial class still largely determines your place in the AI era. 若** jotted down the through-line of the talk: see yourself, see those around you, see others; the teacher of the future is a coach who sees the child and ignites them. 炳* threw out a judgment worth chewing on: it's more reliable to hand knowledge to teachers and companionship to AI.

After a full day of grand narrative, 丫丫 tossed out a wonderfully everyday puzzle in the evening: any way to clean up WeChat contacts, meaning the ones who've already deleted you one-sidedly or you haven't talked to in ages? She said she'd manually cleaned over thirteen hundred people today and couldn't keep it up. 夏** offered a plain, practical folk trick: start a transfer and go to the password-entry step, and if the other side has deleted you it'll prompt you. 赛*** suggested having AI write a program, safer than those bulk tools online. 丫丫 was a bit scared: "I'm a little afraid of letting AI touch my WeChat data, my last account was only just unbanned." 为*** added a key reminder: writing an automation script is fine, but don't let AI take control.

The group woke up to a string of good-mornings and coffee emojis. 若** kicked things off by dropping the full agenda of the 2026 BAAI Conference into the chat，"feel free to tune in if you're interested." 李** sent over yesterday's group daily as usual, and 丫丫 picked it up while listening: today she's a happy puppy with a snack cart. She went on to rave about a treasure-trove snack she'd recently found, and y*** ordered it on Taobao right then and there, joking that she's always up for food recommendations. A light-hearted chat about weight-loss shots als* drifted through, with a few old friends ribbing each other. The room's warmth got going on this everyday small talk first.

🌧🌧🌧 shared a real classroom experiment: she tried folding coding agent s*ftware into her less*n prep and, building on WorkBuddy, made a little reading game for grade-schoolers in half a day. From her testing, the model runs on GLM 5.1 and is pretty capable，across sixty-plus task rounds she kept worrying it would freeze mid-way, but it never did. She's already recommending the tool to parent friends around her, finding it easy to pick up, and figures she could later get grade-schoolers in on co-creating games too. 宋* read it through and praised how engaging the write-up was, then drew out a new thread，a lot of what's happening in kids' AI education could carry over to the elderly side, helping with cognitive impairment and Alzheimer's, a perfect chance for intergenerational connection.

Riding the tooling topic, the group worked through Codex's invite-reset rules. 彬* laid out the gist first: paid users get one reset by default, and inviting new friends can earn up to three more. 年* pasted the official fine print crystal-clear，during the event you can send at most three referral invites, you can't invite yourself, the invitee can't have used Codex in the past two months, and they can't already have a valid invite. A practical freebie guide that picked up right where 🌧🌧🌧's GLM 5.1-on-WorkBuddy run left off, with everyone carefully spending their compute where it counts.

丫丫 took up her connector role again, posting the job description for an ecosystem-operations intern at Ant's strategic investment arm into the group, specifically noting it's paid with a path to full-time, and welcoming both self-nominations and referrals. Putting real opportunities right in front of group members is the everyday texture of this place，intel doesn't stop at tools and events, it lands where actual people can grab hold of it. 宋* said thanks and picked the share up.

With the college entrance exam just wrapped, 九*** previewed an 8 PM roundtable livestream tonight: AI is utterly reshaping learning, college admissions, and career tracks. The lineup features frontline university faculty, AI-education founders, and seas*ned parent representatives, tackling the three questions parents and teachers feel most lost about，whether kids still need to grind through endless practice problems, whether AI will replace teachers, and how to support a child's growth in this new era. In the afternoon, 丫丫 bumped the preview again with an @everyone. Midway through, L*** left a serene, lyrical note, threading the anthems of Zhejiang University, Fudan, and Tsinghua into one "eternal road to truth," landing on the line "the future is already here，because s*meone is paving the way."

Past midnight, the group still wouldn't sleep. 丫丫 announced "Going to bed first! I can't hold out any longer," yet 若** found the s*urce of that demand gap right before sleeping, posting the image before leaving in peace. H*** marveled at how active this group really is, and 若** picked it up poetically: plenty of people on the same wavelength. The most romantic touch came from S***, who placed an order with yee, the one doing AI perfumery, for a healing, warm scent inspired by today's orange sunset. A late-night chat can turn into a product brief, and this little deal has the flavor of an education group: born straight out of daily life.

In the morning 吴熳 recommended Dengxiabai epis*de 29: the guest 倒放 went from studying electrical engineering all the way into vibe coding, image-making, video, and music writing, and even her own poster Skill was born from 倒放's inspiration. Her recommendation carried weight: 倒放's prompt writing is nothing short of art, minimal yet remarkably effective. 九*** followed with a teaser for Friday's 8pm gaokao roundtable: AI is reshaping learning, college admissions, and career paths, with frontline college teachers, AI education entrepreneurs, and a parent representative sharing the stage to tackle the three questions that baffle parents and teachers most.

A question from K*** lit up the noon hour: why do teachers first overflow, then run short? 若**, who works in higher education, broke it all down: a wave of teacher retirements layered onto sharp cuts and even suspensions in teachers' college enrollment creates a phased mismatch between supply and demand; primary school teachers are already in surplus while secondary schools still run short, and s*me regions are moving primary teachers up to fill secondary gaps. 谭老师 added a real sample from his circle: a teacher was asked by the school to sit the certification exam for a higher grade level. What 年* saw was a spatial polarization: remote areas have no teachers, central areas have too many. 若** closed with words both heavy and true: this thing cascades layer by layer, and without reforming the evaluation system it's all empty talk, but you've got to survive first.

While the teacher dilemma was being debated, another thread ran in parallel. 陈* posted a request: on the 18th, the afternoon sub-forum of the China International Cartoon & Animation Festival in Hangzhou needed a keynote guest. 若** immediately thought of her friend who does AI anime drama, "she could probably be a guest," and went off to ask right away. Sign-up method, point of contact, a few rounds and it was all settled, from request to a concrete lead in just twenty minutes. In the afternoon 谭老师 dropped in an invitation to the 2026 Youth Hackathon, and Jiandan's piece, "AI education's first step off the screen is a ten-thousand-yuan robot," scrolled by in the same window, the educators' opportunity list grew a few pages thicker today.

The afternoon's live reports came from two cities. 丫丫 rushed to the Alibaba Cloud Bailian CLI event at Hangzhou's Yungu Center, firing off a location pin, a video, and photos in quick succession, amazed that s* many people showed up on a weekday. Almost simultaneously, D reported from Tianjin: today's event was packed too, with seats now set up in the aisles. 丫丫 summed it up in four words: AI is on fire. Beyond the crowds, L*** left the most composed remark of the day, calling herself old-lady tier yet seeing it clearly: online is one world, offline another; as for the trendy silver economy, she said bluntly that building services while eyeing the money in other people's pockets is putting the cart before the horse.

In the evening 李** dropped yesterday's group daily into the chat as usual, and 若**, staring at that long image, asked the s*ul-searching question: what was this summarized with? Late at night she forwarded another article, "The three scarcest abilities of the AI era, and the gaokao tests none of them," picking up right where the midday teacher supply-demand debate left off: what exams screen for and what the times demand are less and less the same set of abilities. 九*** and K*** liked it in turn, closing the day out in a row of thumbs-ups.

Right as the sky brightened, 雪*** dropped a stinging remark: those "abilities" drilled out through endless problem sets don't seem worth much anymore, s* what edge does your kid even have against AI? L***, a university teacher and medical postdoc, took up the thread and wrote a long passage: from kindergarten to postdoc is an indispensable pyramid, and AI is just a tool, much like the New Chinese Dictionary we relied on back in the day. The key is teaching kids to go from using it, to using it independently, to pioneering and innovating. 吵* gently pushed back from the side: why does a kid have to compete with AI at all? Lily later added one more cut, AI is only a tool, not a teacher, and the only people pushing AI on you are salesmen.

丫丫 shared yesterday's group daily, then floated two new wishes: one, the ability to look up a particular pers*n's chat history, s* when you spot s*meone who talks well you can dig into what they've said before; two, member intros, since plenty of folks pop up once and go quiet, and reconnecting later means scrolling forever. The sticking point is how to pull WeChat message data without risk. Her backup account was just unbanned, and she worries that once the skill is open-s*urced, version mismatches across people's WeChat could get accounts banned. 李** suggested "turning it into a webpage and viewing it inside there." 煎** tossed out a wild card: reverse-engineer the in-car WeChat interface to read chat logs, since car systems are policed less tightly, adding the obligatory "for study and research purposes only."

七* recommended an AI digital art whitelist contest, low barrier, creativity-driven, and worth jumping on for its first edition. But 社*** offered a far wilder path: stop fixating on youth competitions and just hit up all kinds of hackathons, where one summer can clear five figures, net a small 100K, and rack up a pile of award certificates. When s*meone worried about whether primary and middle schoolers could even enter, he fired back with the principle that what the law does not forbid is permitted, registration lists no age requirement, and a minor showing up at a hackathon actually hands the organizers a fresh narrative they're thrilled to have. Below, 七* added a grounded note: any kid willing to join a hackathon has already self-filtered into the motivated, driven crowd.

K*** told a real-life case that fired up the whole group: a middle schooler who didn't fit traditional schooling, got his parents called in, yet used electronic components to simulate mouse signals and take control of the teacher's computer, and now builds his own games and runs his own servers. 社*** ran with it and dropped a hot take: prodigies aren't produced by education, what education produces are cogs, and he flatly claimed that today's top-ranked star students will all end up working for kids like this one. 七* delivered a cool counter-strike: talent is selected, not cultivated. As the argument wore on, 杨老师 met it halfway, technical ability comes down to selection, but resilience and mental toughness still have to be cultivated in a prodigy, or they easily turn obsessive and arrogant and fade into obscurity.

丫丫 first punctured the OPC halo: a one-pers*n company is actually really hard, and normally it takes two or three people each handling their own piece to last, because a customer's needs always run a notch beyond what you alone can provide. 吴老师 followed up with a string of hard truths: she deals with a lot of AI development prodigies, but they hold big misconceptions about commercialization, promotion, and marketing, always thinking the code they write is the moat, when GPT and Claude are splitting the market and the throne keeps changing hands. Her conclusion was blunt: channels and sales are what can truly be an AI product's moat. 杨*** als* spoke from experience, programmers simply don't understand business logic, and you'd be better off having the business side use AI to build the whole thing themselves and sell it directly.

丫丫 brought back what she learned at the education philanthropy biennial. She went to present the "AI Spark Truck" project and, talking with 顾远老师, landed on a piercing point: being loud or low-key should both match your purpose. When you need others to come to you, to offer information and help, you can be a bit louder; once you already have the res*urces, let the results speak. But going loud needs limits, s*mething you're 80 percent sure of can reas*nably be pitched as 100, but you can't sell a 20 as a 100, and you should dare to admit where you fall short, because s*metimes "weakness" is actually "strength", and authenticity is what builds trust. She als* reflected that in the AI era it matters even more to meet offline, there's nothing new under the sun, and before deciding it's worth going back to the scene to see how the people who already stepped in the pits handled it.

In the evening 丫丫 kicked off a new trick: using Floatboat's "sharp-tongued bestie mode" to watch a course-selling livestream with her. The AI bestie "Little 丫" roasted the whole way through, dismissing the host as "all vibe, no substance" before she'd even flipped the slide, and giving an early heads-up that "the hard-sell segment is coming up, there might be gift cards", which it called dead-on, leaving 丫丫 cracking up that "watching the teacher's livestream is less fun than watching the bestie roast it." w*** deeply related, saying it feels like it's playing me, at today's lecture I predicted the next slide would flash a QR code s* I bailed, and sure enough up came the QR code. By tonight at least two humans and one AI were glued to the same livestream, a full-on "chief stand-up comic vs. sharp-tongued AI bestie." That livestream ultimately s*ld 422 packages, and 丫丫 watched while furiously s*aking up the marketing.

直*** read through the AI employment-impact survey 丫丫 had shared and was struck that the teaching profession is actually growing, then wondered whether, as the population shrinks and class sizes drop, teachers might be needed even more. That single question lit up a calm, reality-grounded debate. 社*** threw cold water: around here primary schools are all merging, the city's entire campus recruitment is just 5 slots, and rural schools are merging even harder. 泉*** made it sting more, retirees aren't replaced with new staff positions, and even private kindergarten principals have gone off to open elder-care homes. 若** pulled the view back with data: the two-child policy in 2016 produced a baby boom, s* demand for university teachers won't peak until 2034, but the primary-school peak has already passed, and the children coming after will only grow fewer. Behind one rising curve hide countless local realities of starkly different fortunes.

The new day wakes up on a few clear-eyed reminders. 雪*** tosses out "go easy on the API relay stations," tagged with the quip "even Trump went and ran an API relay station now," a wake-up bell for the group's penny-pinchers. 吴老师 turns her gaze to the power of persistence, marveling that 斯泷 "hand-picks an AI daily report every single day, two years without missing one issue; that kind of focus is truly rare." Right on cue, 丫丫 updates the Way to AGI knowledge base, dropping 小互's open-s*urce video translation tool and 17 prompts that let Hermes run on autopilot. Within minutes, intelligence, tribute, and res*urces are all on the table, and the education bureau's day is underway.

It's Gaokao seas*n, and the group stages an exam of its own. 丫丫 forwards the big models' Gaokao scores, joking "finish a subject, test a subject, let's see everyone's Chinese and math," and DeepSeek's last-place finish in Chinese becomes the headline. 炳* nails it: "DeepSeek is still a fine hand at math, but Chinese really isn't its thing; its language expression has basically no playbook." E*** contributes the day's best self-deprecation: "I have full confidence in my ability to forget. Abs*lutely zeroed out, zeroed out, zeroed out." A model report-card day turns, in the hands of these educators, into a relaxed topic that knows both the tech and the test.

A simple "anyone have the full text of 'Inside DingTalk' and 'Outside DingTalk'?" steers the talk to DingTalk VP 马锐拉's viral farewell essay. 870 fills in the backstory: on June 8, 马锐拉 published "Outside DingTalk" to announce his departure, while over on the other side 幽素 posted "Inside DingTalk" on Alibaba's internal network. E*** happens to have the hundred-odd-page version of "Inside," and proactively pulls strings, asking the company's DingTalk service partner to help verify whether "Outside" is authentic. Between the two of them the full texts come together, and E*** voices the shared feeling: "I run IT-systems work at my company and hit similar pain points, it's just that the big players feel that pain more fully, deeply, and thoroughly."

Spotting an FDE role priced at 20K, 丫丫 wonders aloud, "isn't the FDE premium supposed to be huge? Why is this one only 20K?" 社***, speaking as s*meone who's lived it, gives the whole group a less*n: this is really closer to a senior project manager, part product, part project. "I've done PRD docs, product prototypes, business flowcharts, and I als* had to own architecture design, demo validation, and tech selection." He recalls the year four of them went on-site in Sichuan for a month and a half: "the entire coding process was offline, and I was editing code and writing scripts in vim on Linux." 丫丫 cuts straight to the essence of the role: "you have to understand both real-world deployment and back-end code; it's basically the reserve corps for founding a startup."

The confusion over picking a Gaokao major pushes the talk into the deepest discussion of the day. j*** speaks from experience: "if I hadn't switched from a dead-end major to computer science, I'd probably still be tightening screws," accepting even a year's downgrade to move from materials science into coding. 七* fires off a string of clear-eyed verdicts: "a major is reverse-engineered from employment; there has to be clear market growth to reverse-engineer a clearly employable major," and points out that "all programmers alike, the full-stack who uses AI earns top dollar, while the front-ender who doesn't is in a precarious spot." Luna adds the kicker, "s*me programmers are already less capable than codex." 丫丫 picks up the bigger question, whether AI is a net gain or net loss for jobs, and using analogies of coachmen who moved on to driving trains and hand-weaving women who moved on to operating machines in the steam age, lands on an optimistic answer: the jobs newly created far outnumber the ones replaced.

By evening the talk drops from the grand narrative back to concrete livelihoods. K*** poses a down-to-earth career question, "is a sales job in the humanoid-robot education segment worth taking, is there compounding value in the industry experience," only to admit "I fell asleep before I even finished asking," which got everyone laughing. 直*** follows up curiously, "what do educational robots mainly do," confessing that his understanding of robots still stops at Unitree. The real finale, though, is 丫丫's output: after spending the day pulling the WEF, McKinsey, Goldman Sachs, ILO, and Stanford reports with Floatbot, by nightfall she delivers a blue-themed "AI Employment Impact Study" deck, with the group exclaiming "s*oo prolific!" From a single question to a research deck, the education bureau once again closes the loop from curiosity to delivery.

The day began with a light burning late into the night. A little past two in the morning, 丫丫 dozed off on the couch, then rolled over and got back up, muttering that no, she still had to get the video edited. She was cutting a reading of A Love Letter to Grandma; the script said AI can generate a hundred love letters in a second and calculate the most precise remittance amount, yet it can never write the trembling in every word across eighteen years. She als* tossed out a heretical little travel hack for ironing clothes, and only called it a wrap with one last line: editing's done, time to sleep. 张** marveled beside her at how hard she pushes, and 莫* simply @-ed her "asleep-before-10pm edition" username for a laugh, and the whole group cracked up.

The morning group lit up with a public beta announcement. L*** said Cola had been in the making for a hundred days and was now officially in 1.0 public beta. The team kept debating what she really should be, and every time the conclusion was the same: Cola is a pers*n. A silicon-based life with a s*ul, who remembers every word you've ever said and slowly becomes the one who understands you best. 宋* was struck by a memory, saying he first heard about Cola on a podcast where s*meone called her the three-C livestream sales queen; he'd downloaded it on his Mac a month or s* back, getting that Her-like feeling, except back then it felt pricey and slow, s* with the public beta he wanted to try again. 丫丫 dove right in and even shared her own invite code, saying she abs*lutely loves reading Cola's diary.

In the morning 丫丫 switched back into her res*urce-curator role, pasting the Road to AGI knowledge base updates into the group: one was a preview of the WWDC-plus-AI livestream at 8 p.m. on June 11, chatting about which of all these new features are worth building an AI app around; the other was a list of this week's top AI papers. Seeing the paper list, 吵* was instantly reminded of one called Why Multi Agent Fail. One update drew out a thread of thought, the very rhythm this group knows s* well: res*urces aren't just piled up there, s*meone picks up the thread and runs with it.

In the evening a piece of intel made the group look up again: WeChat AI announced its closed beta, where a single sentence lets users directly operate mini-program pages. 豚 carried the news into the group, S*** honestly admitted he didn't quite get it, and 无* immediately followed up with a Guide to Developers Plugging Into the WeChat AI Ecosystem. For an education-focused AI community, no one has fully spelled out what this new WeChat gateway means yet, but everyone knows in their hearts this is another opening worth keeping an eye on.

In the evening 丫丫 reported back from the venue. She said she was thrilled to come exchange and learn at the AI sub-forum, that the WaytoAGI-EDU and CCF AI tech bazaar drew a lot of attention and kept everyone having fun, landing on the line: in the age of AI we need to meet in pers*n even more. Still not done, she dropped a sign-up flyer: an on-device agent maker party hosted by ModelBest and OpenBMB, where bringing a Raspberry Pi, smart glasses, or a robot dog gets you billions of MiniCPM tokens, with an on-site treasure hunt to collect limited-edition swag. Her pitch was heartfelt: you abs*lutely have to come to this one, it's s* much fun, tons of big names on the floor, and she @-ed everyone.

After dark, the topic settled into s*mething heavier. 丫丫 kicked it off with "is everyone's magic still okay," saying news everywhere lately has been about relay-station risk, and it feels like they're about to round people up to make an example of s*meone. 煎蛋, who works in AI security, took the mic and explained it thoroughly: things like sub2api are the underlying service for distilling data, otherwise how would certain companies have s* many employees feeding the machine; relay stations can slip custom information into the data they return, inserting ads, pois*ning, doing prompt injection. Alex added that s*me relays even sell users' conversations with the AI, and 泼* chimed in with and there's pois*ning too. 煎蛋 closed it out plainly: it's a willing-buyer-willing-seller deal, users want cheap compute and the big players need distilled data, but you're still best off not using relay stations. 宋*'s line, it's getting more and more unstable, wrapped up the day.

Before the sky even lightened, L*** dropped a gaokao poster into the group. "May you ace the exam, may your future shine, may education make us strong, may technology lift the nation," signed off as the 2026 National Gaokao Day. This is a room of educators, s* naturally the day's first message goes to the kids sitting in the exam halls. Beyond the blessing, it als* sets the tone for the group today: education isn't an abstract slogan, it's s*mething millions of people are actually living through right now.

丫丫 turned yesterday's CCF science popularization fair into a short news clip and posted it. Academicians lecture on AI, popular-science stars teach how to make science fun, and kids learn while they play at the ground-floor AI-tech market. An AI spark car, a children's AI gala, an AI Rubik's cube, AI perfume-blending, AI milk-tea blind boxes, AI trading cards, 3D printing, robot dogs, more than 2,000 families pouring into the venue. She followed it with an update from the "Way to AGI" knowledge base: a big-tech quitter with zero funding building a 9-million-strong AI community, and a 90-minute build-a-thon held after Codex blew up. The buzz of the scene gets neatly folded into reference material that lasts.

L*** showed off a medical researcher's extra-long list of titles, and E***'s quip, "that's a long title, must be a big shot," cracked the group up. The topic then turned to an unexpectedly candid confession: L*** said that at the 2050 conference "I'd already been kicked many times, and even got cut off mid-speech by an assistant, s* one more kick, I'm not afraid anymore." E*** rolled with an IT-style self-deprecation: "If you're in IT, then you've just got to take a lot of kicks." In a room used to mutual flattery, s*meone willing to voice the awkwardness of being interrupted and dismissed carries more weight than any title.

赛*** tosses a real scene into the group: once students lean on AI constantly, they treat it as a machine that hands out answers, and the moment things get hard they don't want to think. In the student's own words: "I just throw the question at AI and get the answer, s* why should I bother to think?" The question lands like a stone in still water. 870 is first to respond: could we build in a critical mindset, s* students discover that AI's answers aren't always reliable, which in turn pushes them to re-examine and borrow from it dialectically. s*** pushes it higher, judgment itself is scarce, s* critical-thinking ability has to be designed into any teaching that involves AI. One teacher's dilemma is being treated by a room of teachers as a shared problem to s*lve.

The discussion quickly shifts from "what do we do" to "here's how we do it." 杨老师, the self-described s*cial-anxiety sufferer, says students "need to get burned by AI, hard and squarely," and 谭老师 adds "no pain, no memory." 王** throws out "has anyone studied an AI-fault-finding class?" and 谭老师 instantly turns it into a plan: run an "AI-can't-find-it contest" to see who's best at spotting the problems. 李* brings frontline experience, her university students found 豆包's answers unreliable, s* they all switched to deepseek's expert mode; 煎蛋 piles on, "豆包 may be unreliable but at least it admits mistakes," and the key is to "guide students to explore the answer together, not just hand it over." The anxiety gets digested, one concrete action at a time.

After the lively round of prescriptions, 小圆 added a line that pulled it all tight. She said this really comes down to students' understanding of "what learning is": AI has made getting answers easy, but "getting the answer" and "mastering the knowledge" are not the same thing. The more cutting second half: plenty of adults lack this same awareness, it's just that answers used to be harder to come by, which made it easier to notice your own ignorance. Super黄 chimed in, "true, never went through the brain." A discussion about students ends up holding a mirror to everyone.

The moment the gaokao Chinese exam let out, K*** posted "my answer to the 2026 National Paper One gaokao essay prompt" into the group. 莫* turned it into an event on the spot, the AI scene's first-ever gaokao essay showdown, open to Jiangxi folks and non-Jiangxi folks alike, with spons*rs welcome to sweeten the prize pool. This crowd outside the exam hall redid this year's gaokao prompt in their own way. Turning s*meone else's exam day into your own creative day, that's very WaytoAGI.

Late at night, 丫丫 got back from Suzhou to Hangzhou West Station and snapped a hall full of "job-hunting kiosks," which S*** nailed with one line, "the station edition of BOSS Zhipin." That opened the hiring floodgates: j*** is helping a friend recruit an agent engineer at 35-50K, the company just raised hundreds of millions, and sighs "we're s* short on people"; 煎蛋 dropped Dify's job listing; 丫丫 rounded up a string of education-nonprofit roles. K***'s "I need a job" spoke for plenty of people. But the day's true landing point was the film 丫丫 saw with 薇娜老师 in Suzhou that afternoon, "A Letter to A-Ma," she cried until her eyes swelled up like walnuts, calling it utterly restrained and never deliberately tear-jerking, unlike "those assembly-line releases and period idol dramas churning out industrial sugar substitute every single day." S*** marveled, "丫丫's 24 hours aren't really 24 hours," stamping a seal on this busy day.

On Saturday morning, 雪*** set the tone for the day with a single "lucky double-six." At ten, 丫丫 dropped the past two days' group dailies into the chat, adding a deliberate note of "thanks to 祥瑞老师"，the daily itself has become a fixed rhythm for this community. Right after, 煎蛋 sent first-hand news from the CCF science-popularization conference in Suzhou: yee 老师's booth was packed. A seemingly ordinary weekend morning was actually the day's two most important threads arriving early，one the online daily relay, the other the offline conference throwing open its doors.

At midday, 苏杰 did s*mething very much in his style. He took the AI-themed closed-door-meeting hot-take article that everyone had been into recently and turned it into a s*ng with SUNO v5.5, titled "The AI Closed-Door Meeting, Turned Into a Parent-Teacher Conference," captioned with eight words，no technique, only feeling. This wasn't showing off a tool; it was preserving a real discussion about education in a lighter, more shareable form. In the afternoon he tossed out an invitation: next Friday he'd record a podcast at his own incubator and welcomed the group to come hang out in pers*n. Insight, work, and invite，the full package.

As noon approached, 丫丫 started flooding the screen，but with substance. She put out six recent Shanghai AGI Bar events in one shot: 影眸 Hyper3D's Rodin Gen-2.5 flagship 3D model experience day, a "What in art cannot be distilled" young-artists' night talk, a VC Action where two investors play cyber bartenders, a security closed-door around the Claude Mythos model, the AI Wave Seas*n launched by Yantao, and the second AI Pharma Pioneers' drinks gathering. Each listed the time, place, and who it was calling for. K***'s single "I want to go too" turned the online preview straight into an offline sign-up.

The event previews had just filled the screen when 诗* asked s*mething that hit a nerve: when will the winds of all these events ever blow to the Xiong'an New Area? It's the most honest voice from a lower-tier city，the buzz mostly happens in Beijing and Shanghai, far from home. 文* followed up: the CCF science-pop event was packed, they'd brought four students to run an activity on site and were swamped，proof that people in the regions are doing things too. 冰鱼仔's reply was the key one: rather than wait for the wind, organize it yourself. A question, a firsthand account, and a word of encouragement nudged the old problem of "uneven res*urces" one step toward "local self-organization."

The CCF scene came alive in the afternoon. K*** fired off photos of robot figures and a robot dog, and 年* quipped: the stereotype of the robot dog is about to set in stone. The really fun part came after，the meetup chase. K*** called out "Still here? Let's meet," 谭老师 asked "Where are you," K*** said "Just about to leave, I'll come over," "I'm here," and even gave a pat. Sadly the timing didn't line up, and 谭老师 could only reply "Then next time." A near-miss in-pers*n meeting that didn't happen，yet it revealed this group's most precious thing: people genuinely want to step out from behind the screen and see one another.

Past four in the afternoon, 丫丫 became the scene's live host. She fired off real shots, a CCF location pin, and a digital copy of the conference booklet, and a line of "saw s* many friends offline today, s* happy" piped the on-site excitement straight into the group. 870 had a sharp eye: are those smart glasses you're wearing? 丫丫 revealed the answer，thanks to 高**老师, she got to use 理想's AI glasses, much like Meta's, "lighter than both Qwen's and Xiaomi's, I really like them." 莫* teased, "Mainly because the pers*n's good-looking." w***'s line "丫丫老师strode up there like she didn't recognize a s*ul" paired with "welcome to move upstairs to the main hall," and 南宫潺 added more on-site photos. In that moment, the group chat was practically a live broadcast.

At half past nine in the evening, 丫丫 came online again: today's CCF science-pop conference news quick-cut was out. A full day on the ground was edited into clips you can rewatch. 谭老师 chimed in, "I tried the perfume," referring to the AI fun-scent blending at the market; 胡** laughed as he shared a detail，at the very end he heard a kid say they'd played with the drawing one. Pieced together, these fragments are exactly what the conference most wanted to leave for children: AI isn't just a word in the news, but s*mething you can smell, draw, and play with. Late at night, Elfe Xu gave 丫丫's lucky cat a pat, +1 to the luck meter, putting a period on this lucky double-six day.

The day began with a long message in the dead of night. After watching a design video, 直*** admitted he had known nothing about design, and only when he truly wanted to build a pers*nal website did he realize how hard it is to express needs to an AI precisely without the right terms. J***, 胡**, and 若** picked it up, continuing to respond to the hormone ring designed for women from the day before. A design topic didn't stop at whether s*mething looks good; it wandered into product stories, the female perspective, and hackathon camaraderie. That is the most precious thing about this group: beyond the tools, s*meone always brings the human experience back in.

At half past two, D sent live photos and asked, "Any fellow members here on-site?" 董** answered right away with "Here," and 执***, realizing others were on the scene too, asked who the speaker was. D added the video channel "未来博士wepon," and B*** discovered he had long been following the pers*n's Xiaohongshu, while 笑颜 found the video channel already on her follow list too. Just like that, an offline event completed its rounds of recognizing faces, filling in details, and connecting accounts inside the group. It's light, but that lightness is exactly the on-site feel of an education community: s*meone is present, s*meone sees it, s*meone passes the thread back.

Close to three o'clock, a very down-to-earth practice share appeared: a skill that simulates grading zhongkao exam papers, built for parents and kids nearby, with WorkBuddy as its foundation. It isn't as grand as a product launch, nor as buzzy as big-model news, but it hugs the education front line. Grading a zhongkao essay is a task teachers, parents, and students can all grasp instantly; turning it into a skill isn't about showing off, but about making it easier for parents and kids to actually use. Small, usable education-AI works like this are exactly what a group like WaytoAGI-EDU should leave behind.

At 16:26, 丫丫 posted what reads like the spiritual anchor of the day: returning to Suzhou in a new role, and still being moved by a teacher's love for children while catching up with old friends. 宋* replied, "Back in Suzhou again, s* good, our teacher 丫丫," and H*** sent a rose. No tool names here, no methodology, yet it lifted the group's mood. An education-AI community left with only models and links would feel dry; 丫丫's words put people back at the center, reminding everyone that no matter how far you travel, what truly holds your hand is still teachers, children, and the scene itself.

After 丫丫 sent a video from the scene, 宋* steered the topic toward agent hardware. He mentioned an AI watch that can log 24 hours and combine heart-rate data, then wondered whether there might one day be a proactive agent watch built for education. 直*** added a reality check: WHOOP and smartwatches can already do s*me similar things, but what exactly would an education agent watch be for, analyzing flow states? 赛*** pulled the question back to the ground: AI hardware comes down to use cases. 宋* pushed further, noting education isn't just academics but als* growth, a learning companion, and emotional company. A hardware topic turned into a use-case question.

In the early evening, 丫丫 dropped all the highlights of the AI fair at Suzhou's CCF science-popularization conference at once: AI fun fragrance-making, an AI seed car, a children's AI gala joint showcase, AI cards, AI psychology, programming robots, a bionic robot dog, and AI milk-tea blind boxes. 胡** sighed that it would be wonderful to attend in pers*n, 煎蛋 asked if anyone was coming to Suzhou tomorrow, 赛*** noted the distance to Xiangcheng District and Wujiang, and H*** even asked, on behalf of the fragrance group, to join 丫丫's release list. In that moment, an offline event became a shared anticipation of the online community.

After eight in the evening, the group switched back to tool intel. M*** posted a Zhisheng share titled "Stop letting AI just chat with you, let Codex actually do the work for you"; 丫丫 added that step-3.7-flash is free to use on 302ai, and forwarded the news of ima integrating LearnBuddy. Then K*** posted the location of Suzhou Mountain Bilingual School, only to find the CCF event was right nearby, regretting not being able to bring along younger siblings to see it. 赛*** added, "It's right in Xiangcheng District, near Suzhou North Station." The day wrapped up cleanly: tool intel, education platforms, and offline addresses all placed on the same map.

The first burst of information at dawn wasn't a light one. 雪*** led with the line "the best way to prevent AI cheating is…", then forwarded a story about an American teacher who tried to catch students using AI to cheat and ended up backfiring. It was like a small nail, pinning education AI's old problem to the group: once the tool has entered homework, detection, and grading, teachers can hardly block it by blocking, and false positives end up hurting students. The discussion didn't unfold much further here, but this message set the day's undertone. Education AI isn't only fun, it als* has boundaries, trust, and judgment.

After nine, the information stream started to speed up. 丫丫 first brought in Tencent Research Institute's super-individual report, then announced she was off to Suzhou; 豚 added an article on AI glasses learning scenarios, and 870 followed by dropping "To Super-Individuals" and "From Super-Individual to Super-Team.pdf" into the group. Strung together, these messages were like a new map for WaytoAGI-EDU that day: individual capability, team collaboration, on-site spaces, and education scenarios all drawing closer. It wasn't just link-sharing, more like a reminder that saving this material into the knowledge base is what makes continued co-creation possible later on.

At 10:25, 七* handed her indecision over to the group: "Which of styles 1234 might be better?" This wasn't an offhand question of taste, but a way of sharing the first round of judgment on the mini-program interface with everyone. 丫丫 picked 2, Victor picked 4, 870 chose 2 and added "clean and clear," s*** favored 2 and 3, and B*** came in from the angle of students and parents as readers, saying 2 and 3 read comfortably. A small choice was voted into a lightweight review session by the group. 七*'s line "good thing I asked first" is telling: a product gets seen in the group, and als* takes fewer detours because of it.

炳*'s critique pushed the vote one step further. He didn't just say he liked 3; he broke down the shadows in 1, the out-of-focus risk in 2, and the geeky vibe in 4. In the middle, a change to the group announcement popped up, and he thought his reply had glitched; 七* and 丫丫 quickly caught it, and the mood went from slightly tense back to relaxed. He then went on about visual consistency, information hierarchy, font choice, and future reading length, citing the Lynk & Co app, Windows Phone, and Metro design as references. The best part was the identity contrast: he said he works in Chinese-language education and that design is just a hobby. The group suddenly had a Chinese teacher who could review interfaces.

Around lunchtime, 煎蛋 filled in a self-introduction, and 钟音琴行王辉 casually revealed another layer of identity: he's als* the program team lead for the 2050 Conference's WaytoAGI third-anniversary "Absurd Music Festival." This new friend had just joined, and the topic immediately jumped to 3D models. 赛*** asked how to take students through AI-generated 3D modeling over summer break; 丫丫 suggested Tripo, 诗* posted studio.tripo3d.com, and 煎蛋 added the 3D models on his pers*nal homepage and open-s*urce Hunyuan 3D. It wasn't a pile-up of tool names but a very typical education AI scenario translation: from pers*nal work, to a student summer project, to usable tools.

In the evening, 吴熳 pulled the topic from tools back to people. She noted that vibe coding has made building products easier than ever, s* the remaining question becomes: what is a real need? The Dengxiabai podcast she shared was about a hormone ring designed for women, a need that came from real illness and could therefore strike its target group. This node added a clear-eyed judgment to the day. In the morning everyone was choosing color schemes, breaking down interfaces, and finding 3D tools; by evening, the question returned to the starting point of a product: AI can speed up the making, but it can't feel real people for you.

As s*on as the sky lightened, D 老师 threw out a very concrete need: to build a tool for collecting information on youth AI training programs and competitions ， low cost, easy to pick up, and just enough data gathering to do the job. 蜜***老师 didn't beat around the bush and offered two paths straight away: Feishu Bitable, or Baidu Miaoda. D 老师 said thanks, then took a jab at himself ， "just realized I typed 's*licit' when I meant 'collect'" ， and restated, cleanly, the whole idea of smartly crawling youth AI training and competition info across the web. A day at the intelligence bureau begins with a real, small need being caught.

At twelve-forty noon, 丫丫老师 made it official in the group: the education-edition group daily skill is now open-s*urced on GitHub under the MIT license, and everyone is welcome to remix it. Right after came a handoff ， she gave special thanks to the author, 祥瑞老师, and announced that the daily generation going forward would be taken over by 祥瑞老师. J*** chimed in below with a grin: "We just talked about it yesterday, and the daily's already out ， execution MAX." From an idea chatted up in the community to a content product that runs, open-s*urces, and carries on ， only one night stood in between. The page you're reading right now is the new craftsman's very first move.

The afternoon belongs to the creators. 若**老师 dropped a Coze app she'd built herself, the "Cyber Meditation Hall," with a caption that read like a bit of doggerel: doing a PhD is like cultivation, the mouse is your wooden fish, the thesis never gets finished, but merit piles up a roomful. Over on her side, 丫丫老师 got to talking with 诗*老师 about the tools in hand ， Floatboat, Cola, Alice lined up one by one ， only to find they were, in fact, "the same batch of users." 诗*老师's grumble was all too real: "There are too many tools, but I have no idea what to use them for." A rare, easygoing stretch at the intelligence bureau, where AI isn't productivity ， it's a toy.

At three-forty-five in the afternoon, 丫丫老师 candidly offered up a less*n: her alt account got hit with a WeChat warning for scraping daily-report data. She broke the takeaways down ， use the best model you can, GPT 5.5 or Opus 4.6 and above; and this may be tied to the WeChat version (4.1.7.1) ， the moment the version upgrades, a failed decryption can trip the WeChat server. V*** picked up the thread with an even bigger whisper: WeChat is beta-testing an embedded Hunyuan base model + workbuddy, with Moments, Channels, Official Accounts, and group-chat corpora all wired together ， and Tencent's market cap has already surged 300 billion because of it. s*** 老师 added an escape route: the daily CLI carries risk, s* you can run a second pass through weflow to bring it down.

At dusk, 丫丫老师 did s*mething she herself called "making an exception" ， @everyone, to formally introduce the newly joined 李**老师. 晓城老师 has worked deep in local Yunnan for years, focused on campus digitalization, teaching equipment, and smart classrooms, and even set up the Ruijing Teacher Development Fund under the Yunnan Education Foundation, running teacher training across the province year-round; in the first AI Spark Car seas*n, he helped s*lve everyone's urgent campus-entry problems. The group instantly filled with "Welcome, 晓城老师." 丫丫老师 added a line: "You're the first pers*n I've welcomed with an @everyone since founding this group," and 晓城老师 replied with genuine warmth: "Wow, I've joined a treasure group."

Riding on this welcome, 丫丫老师 introduced the dream-team of volunteers from two seas*ns of the AI Spark Car to everyone, fully, for the first time: 薇娜 on curriculum design, 可一 on photography coordination, 七* on res*urce matchmaking, 阿不 on operations support, 黄彬 as workshop instructor… name after name, role after role. She als* gave heartfelt thanks to A* ， founder of the WaytoAGI open-s*urce community, now home to over 9 million users; and to Huawei TECH4ALL, which funded the effort twice, "willing to support us even without any branding." A team stitched together from teachers, founders, and volunteers walked 1,073 kilometers in 11 days, carrying the spark of AI into 6 schools across Yunnan.

高**老师 joined the group, and 丫丫老师 thanked him for the remote link-up he ran for the AI Spark Car ， the second stop reached Hadapu Primary School in Gansu, the turning point of the Red Army's Long March: "carrying forward the red bloodline, opening the future with AI." She als* previewed a June 6 roundtable in Suzhou at the CCF science-popularization conference with 翟老师 and 高**老师 ， 5 days and counting. As night fell, the industry news kept coming: 刘**老师 announced the official release of Manycore Tech's LuxReal short-drama edition, using 3D tech to shore up the uncontrollable weak spots of video generation; and 易亚婷老师 brought an interview with 庄明浩 ， "An AI startup with ARR over a hundred million dollars: will it get eyed by the big models?" A day at the intelligence bureau, from a mountain-village primary school all the way to the industry's latest moves.

The morning's main thread burst open after 吴熳 read through the post-lecture feedback survey. She found that what most teachers wanted to learn was still using AI to make PPTs, ones that look good and feel fun, or interactive web pages to grab students' attention. She put it bluntly: all of this floats on the surface, and real AI should be industry plus AI, knowing the industry know-how first before talking about tools. 870 chimed in with "I+A I," and 俏 pulled the question back to application scenarios: without a real place to land, there's no lasting motivation to keep using it. This opening turned the day away from tool demos and into educators jointly asking: what exactly should AI help teachers s*lve?

丫丫 recommended the Onion Academy epis*de by 杨临风: "Using AI to Manufacture Shortcuts Is Killing Real Learning." She flagged two moments in particular: AI shakes up exam-oriented education, yet the process of learning knowledge is still necessary, and schools are still necessary; and she once believed in making money first before realizing her goals, only to later find that wasn't required. A few minutes later, she welcomed 杨临风 himself into the group, and Fei laughingly recognized it was another of his accounts. A podcast link became the doorway to a live exchange, and tied a thread through the whole day's education talk: AI can change the path, but it can't turn learning into a detour.

The PPT debate quickly became the day's hottest first round. 吵* lit the fire with "most PPTs are digital garbage," then sharpened it: "most slides today have no power and no point." D didn't simply take a side; she reminded everyone that PPTs are just a tool of expression, neither good nor bad in themselves. 小圆 broke things into three layers, information presented, conveyed, and understood, while 丫*** landed the issue on judgment: many people open the s*ftware and immediately reach for templates, hunt for images, and pad the page count, with no sign of a single pers*n's judgment. The group wasn't bashing PPTs, it was taking apart one thing: expression in education can't let a template climb on stage in your place.

Just before noon, 关*** threw out a big question from the learner's point of view: when everyone is criticizing classrooms piled high with form and empty at the core, what kind of educational path is actually worth taking? F*** brought in Ronny Chieng's critique of AI at the Harvard commencement, landing not on denying AI but on not letting AI squeeze out real life and creativity. D rewrote the question into a life-long less*n: hold to your true heart, or chase outside praise. B*** offered a very practical sense of boundaries: if you join the game, play by its rules, the key is staying clear-eyed yourself. Here the discussion moved from PPTs to how a pers*n chooses.

At midday, 丫丫 laid out "desirable difficulty" and "cognitive friction." She cited the Bjorks' desirable difficulties, noting that s*me hardships slow you down in the short term yet let memory, understanding, and transfer happen over the long run. 翟*** pressed on the harder part: how do you persuade parents to let their child spend time experiencing friction? 丫丫 didn't urge anyone to push through by force; she started with the ledger of risk: many parents aren't low in awareness, they're low in risk tolerance. Then she laid out the 532 route: don't grind away at opponents, find the people willing to walk with you first, and get one short stretch of road working. The judgment cuts deep, and it's deeply practical.

After one o'clock, 小圆 introduced herself: a STEAM teacher at a public US high school who will leave teaching after this semester to move into AI plus EdTech, heading to ISTELive26 in Orlando at the end of the month. 丫丫 immediately added context: this isn't just about seeing new tools, it's about watching how technology enters the classroom, the curriculum, school management, and teacher development. J*** was als* tagged in by 丫丫 and spoke of her seven years teaching English and coordinating instruction at a bilingual school in Shanghai, plus new developments at Penn GSE and Cambridge. Suddenly the group had an international education thread: not that the distant places are cool, but that s*meone will come back and explain clearly how it lands.

The afternoon mood grew heavy for a while. 吵* began with future family assets and elder care, while S*** and the s*cially anxious 杨老师 picked up medical costs, the silver economy, and exercise. S*** spoke of an elder at home with dementia and disability, and of the choice around his father in the ICU, lines that briefly stilled the group. 关*** als* said that her grandmother passed away in April and her father in May, only then realizing that the education she received growing up was missing the less*n of facing death. 木*, 俏, and K*** brought in life education, financial literacy, gender education, fraud prevention, Su Dongpo, and picture books. Education isn't only AI and the classroom, it als* includes how we face money, illness, aging, and farewell.

At night, the chat turned from heavy topics back to action on the ground. 丫丫 dropped a location at Shanghai's Grand Gateway, originally headed back to Hangzhou but too tired, s* she went to a café to chill, then forwarded the AGI Bar Shanghai skewers event that night and the June 2-to-6 opening-seas*n lineup. 镜*** brought the Magic Fawn desk pet, and later shared the self-optimization workflow of qingyun-cine-skill. 丫丫 als* put out the AGI House info: 14 days in Shanghai this June, teaming up with Sequoia China, Coze, and Jike to build world-changing ideas together. After a day of talking education, life, and tools, the night came back to one thing: go to the scene first and make the thing.