[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"zeug-en\u002Fzeug\u002Fllms-fuer-pi-coding-agent":3,"related-en\u002Fzeug\u002Fllms-fuer-pi-coding-agent":421},{"id":4,"title":5,"badge":6,"body":7,"category":393,"date":394,"description":395,"draft":396,"extension":397,"image":398,"link":399,"linkText":400,"meta":401,"navigation":402,"path":403,"pinned":396,"publishTime":404,"seo":405,"stem":406,"tags":407,"verdict":418,"visual":419,"__hash__":420},"zeug_en\u002Fzeug\u002Fllms-fuer-pi-coding-agent.md","My LLMs for Pi: GPT, Claude, GLM, Kimi and DeepSeek","Heißer Tipp",{"type":8,"value":9,"toc":382},"minimark",[10,14,17,20,23,26,29,38,41,43,46,48,51,56,59,67,69,72,75,78,81,84,88,96,99,102,104,107,110,123,125,128,130,137,139,142,144,147,149,152,154,157,159,162,164,167,169,172,176,187,189,192,194,197,199,202,204,207,209,212,215,218,220,223,226,229,231,234,237,241,244,247,250,252,274,276,279,282,285,289,292,295,297,303,306,309,312,315,317,323,326,329,332,336,339,342,344,347,350,353,356,358,361,364,367,369,372,374,377,379],[11,12,13],"p",{},"With coding agents, I like talking about harnesses.",[11,15,16],{},"Terminal. Tools. Sessions. Context. Permissions. Extensions.",[11,18,19],{},"All those things that turn a chat window into a useful tool.",[21,22],"spacer",{},[11,24,25],{},"But at some point, a pretty simple question remains:",[11,27,28],{},"Which brain are you actually putting into that harness?",[11,30,31,32,37],{},"I already covered the harness part in the article about ",[33,34,36],"a",{"href":35},"\u002Fen\u002Fzeug\u002Fpi-coding-agent","Pi as a coding agent",". This text is the less philosophical and slightly more expensive part.",[11,39,40],{},"The models.",[21,42],{},[11,44,45],{},"And because LLM prices, limits and model names change faster than I can say “just a quick refactor”: this is not an eternal ranking.",[21,47],{},[11,49,50],{},"It is my state of things in June 2026.",[52,53,55],"h2",{"id":54},"why-the-model-question-feels-calmer-in-pi","Why the model question feels calmer in Pi",[11,57,58],{},"Pi is pleasantly unromantic about models for me.",[11,60,61,62,66],{},"I do not have to change my entire workflow just because I want to try a different LLM. I open ",[63,64,65],"code",{},"\u002Fmodel",", choose a model and keep working.",[21,68],{},[11,70,71],{},"That sounds banal.",[11,73,74],{},"But it is one of the reasons I like Pi so much.",[11,76,77],{},"A model is not the entire identity of my setup. It is a component. An important, expensive, sometimes astonishingly annoying component. But still a component.",[11,79,80],{},"And Pi treats it exactly like that.",[11,82,83],{},"OpenAI through Codex, Anthropic through API, DeepSeek, Z.AI, Kimi, OpenRouter, OpenCode Go and custom providers can all land in the same workflow. Not always with exactly the same strengths, not always with exactly the same limits. But without that absurd tool switch where you start from zero again every time.",[52,85,87],{"id":86},"my-default-gpt-55","My default: GPT-5.5",[11,89,90,91,95],{},"If I could choose only one model, my default right now would be ",[92,93,94],"strong",{},"GPT-5.5",".",[11,97,98],{},"Not for everything.",[11,100,101],{},"But for a lot.",[21,103],{},[11,105,106],{},"OpenAI positions GPT-5.5 pretty clearly for coding and professional work.",[11,108,109],{},"You can feel that too.",[11,111,112,115,116,122],{},[113,114],"br",{},"\nMy personal opinion has been for months that 5.4\u002F5.5 are the \"best\" models, and apparently there are now more and more ",[33,117,121],{"href":118,"rel":119},"https:\u002F\u002Fdeepswe.datacurve.ai",[120],"nofollow","benchmarks"," that support me there 🎉",[21,124],{},[11,126,127],{},"It is strong at refactors, debugging, codebase navigation, planning, sober execution and those sessions where the model should please not forget after three tool calls why it is here in the first place.",[21,129],{},[11,131,132,133,136],{},"On top of that, GPT-5.5 brings more than one million tokens of context, ",[63,134,135],{},"xhigh"," reasoning and very large outputs. For agent work, that is not a small side note. Long contexts are not automatically smart, but they often prevent a model from acting every five minutes as if it had just met the project.",[21,138],{},[11,140,141],{},"The actual sweet spot for me is Codex.",[21,143],{},[11,145,146],{},"The Codex subscriptions are surprisingly generous if you regularly work with coding agents. Pure API usage can get very real very quickly. A subscription often feels calmer there, at least if you actually use it.",[21,148],{},[11,150,151],{},"And this is the important difference to Claude: you are also officially allowed to use the ChatGPT or Codex subscription for that.",[21,153],{},[11,155,156],{},"Pi directly supports login for OpenAI Codex subscriptions instead of making you dance around with questionable token workarounds.",[21,158],{},[11,160,161],{},"That makes a big difference to me.",[21,163],{},[11,165,166],{},"The caveat: frontend.",[21,168],{},[11,170,171],{},"For visual work, layout, courage, taste and the question of whether a page at the end again looks like a very polite SaaS landing page, GPT-5.5 is not always my first grab. It can do that. But I trust it less blindly there than with backend, tooling or logic.",[52,173,175],{"id":174},"claude-sonnet-and-opus-strong-but-expensive","Claude Sonnet and Opus: strong, but expensive",[11,177,178,179,182,183,186],{},"Of course ",[92,180,181],{},"Claude Sonnet"," and ",[92,184,185],{},"Claude Opus"," belong on this list.",[21,188],{},[11,190,191],{},"Anything else would be nonsense.",[21,193],{},[11,195,196],{},"Sonnet is often the sensible Claude default: fast enough, strong enough, good at code, good at explanation, pleasant on longer tasks.",[21,198],{},[11,200,201],{},"Opus is the variant I want to take when a task is really long, branched and potentially expensive because mistakes there would be even more expensive.",[21,203],{},[11,205,206],{},"Anthropic describes Opus exactly in that corner: complex reasoning, long-horizon agentic coding, a lot of autonomy. The newer Sonnet and Opus versions also move in the area of 1M context, which is obviously tempting for large codebases and longer agent sessions.",[21,208],{},[11,210,211],{},"My problem is not the quality.",[11,213,214],{},"My problem is the price and the practical usability in a small subscription.",[11,216,217],{},"And important: a normal Claude subscription is not a Pi ticket for me.",[21,219],{},[11,221,222],{},"With a Claude subscription, I would not use the Claude models in Pi. Not because there are no technical workarounds. There are. But that is not the clean allowed path, and if Anthropic evaluates that as subscription circumvention, the worst case is a suspended Claude account.",[11,224,225],{},"That would be far too stupid for me for a few cheaper agent runs.",[11,227,228],{},"If Claude in Pi, then for me only through clean API access or an explicitly allowed provider.",[21,230],{},[11,232,233],{},"If I am only playing around a little, that is irrelevant. But if I seriously let agents loose on projects, tokens suddenly stop being abstract. Then they are no longer “a bit of usage”, but a bill with personality.",[11,235,236],{},"Claude therefore remains a model I respect a lot, but do not always burn through casually.",[52,238,240],{"id":239},"glm-52-kimi-k27-and-deepseek-v4-pro","GLM-5.2, Kimi K2.7 and DeepSeek V4 Pro",[11,242,243],{},"This is the corner where it gets really interesting for me.",[11,245,246],{},"Not because these models are always better than GPT or Claude.",[11,248,249],{},"But because they are often good enough to very good while smelling much less like luxury panic.",[21,251],{},[253,254,255,262,268],"ul",{},[256,257,258,261],"li",{},[92,259,260],{},"GLM-5.2 is the underrated worker."," Z.AI describes it as a model for agentic engineering and long-horizon tasks. 200K context, large outputs, tool use, coding focus. Especially in the old GLM Coding Plan, this was a small cheat code for me: a lot of model for surprisingly little money. Unfortunately, exactly that plan has recently become significantly more expensive. The model still remains strong.",[256,263,264,267],{},[92,265,266],{},"Kimi K2.7 is the pleasant OSS lane."," Moonshot has released the weights, the model is natively multimodal, has 262K context and is clearly aimed at coding, long-horizon execution and agent work. It is not my model for every delicate production rebuild. But for cheap, good agent runs, it is far too strong to ignore.",[256,269,270,273],{},[92,271,272],{},"DeepSeek V4 Pro is the cost drawer that is surprisingly often worth opening."," 1M context, Mixture-of-Experts with 1.6T total parameters and 49B active, strong coding and reasoning orientation. Depending on route and provider, it is extremely cheap. For small things, cleanup, searching, explaining and agentic work where I do not want to count every token internally, that is very pleasant.",[21,275],{},[11,277,278],{},"DeepSeek is also amusingly direct: the official DeepSeek docs even have a Pi integration. That naturally earns bonus points on my very scientific sympathy scale.",[11,280,281],{},"Very scientific here means: not at all.",[11,283,284],{},"But still.",[52,286,288],{"id":287},"openrouter-opencode-go-and-the-model-zoo","OpenRouter, OpenCode Go and the model zoo",[11,290,291],{},"The point is not that I need five models every day.",[11,293,294],{},"The point is that I can try them without dismantling my setup.",[21,296],{},[11,298,299,302],{},[92,300,301],{},"OpenRouter"," is almost dangerously convenient for that. One API key, hundreds of models, different providers, fallbacks, routing by price, throughput or availability. If I want to know whether a new model works in my real workflow, that is often the fastest way.",[11,304,305],{},"Not in a benchmark.",[11,307,308],{},"In my project.",[11,310,311],{},"With my files.",[11,313,314],{},"With my badly named variables.",[21,316],{},[11,318,319,322],{},[92,320,321],{},"OpenCode Go"," is the other practical lane. It is a cheap subscription for open coding models, currently with models like GLM, Kimi, MiniMax, Qwen and DeepSeek. Do not understand it as a magical everything-flat. Limits remain limits. But as a playground for strong open models, it is very decent.",[11,324,325],{},"And the best part: both are not exotic in Pi.",[11,327,328],{},"OpenRouter and OpenCode Go sit as providers in the Pi model cosmos. Add direct providers like DeepSeek, Z.AI and Kimi For Coding. If a model annoys me, I switch. If it surprises me, it stays in the rotation.",[11,330,331],{},"That is how model hopping should be.",[52,333,335],{"id":334},"would-i-recommend-this-setup","Would I recommend this setup?",[11,337,338],{},"Yes, but not as a fixed best-of list.",[11,340,341],{},"More as an attitude.",[21,343],{},[11,345,346],{},"If I want a strong default, I take GPT-5.5.",[11,348,349],{},"If a task is really difficult and cost is not the first pain point, I look at Claude Sonnet or Opus, but only through clean access, not through Claude subscription workarounds.",[11,351,352],{},"If I want to try a lot, work agentically or be cheaper on the road, GLM-5.2, Kimi K2.7 and DeepSeek V4 Pro are much more interesting than the western model conversation sometimes gives them credit for.",[11,354,355],{},"And if I do not know what is good right now, I take OpenRouter or OpenCode Go and just try it.",[21,357],{},[11,359,360],{},"Not theoretically.",[11,362,363],{},"In Pi.",[11,365,366],{},"With real code.",[21,368],{},[11,370,371],{},"That is the real strength for me: I do not have to religiously decide on one model.",[21,373],{},[11,375,376],{},"I can just work.",[21,378],{},[11,380,381],{},"And sometimes exactly that is the best model strategy.",{"title":383,"searchDepth":384,"depth":384,"links":385},"",3,[386,388,389,390,391,392],{"id":54,"depth":387,"text":55},2,{"id":86,"depth":387,"text":87},{"id":174,"depth":387,"text":175},{"id":239,"depth":387,"text":240},{"id":287,"depth":387,"text":288},{"id":334,"depth":387,"text":335},"Software","2026-06-04","Which LLMs I currently like using in Pi for coding agents, why GPT-5.5 is my default and where OpenRouter or OpenCode Go become useful.",false,"md","\u002Fimages\u002Farticles\u002Fllms-fuer-pi-coding-agent-cover.webp","https:\u002F\u002Fz.ai\u002Fsubscribe?ic=H42QOGJABT",null,{},true,"\u002Fzeug\u002Fllms-fuer-pi-coding-agent","12:15",{"title":5,"description":395},"zeug\u002Fllms-fuer-pi-coding-agent",[408,409,410,411,412,413,414,415,416,417],"llm","coding agent","pi","openrouter","opencode go","gpt-5.5","claude","glm-5.2","kimi k2.7","deepseek","Not every model has to be my favorite model. The strong part is that Pi makes switching between them almost boring.","rings","11B1-QGfOpaNiSMIyIXATjU3kELrnKLhJDNneBrAGdM",[422,611,924],{"id":423,"title":424,"badge":425,"body":426,"category":589,"date":590,"description":591,"draft":396,"extension":397,"image":592,"link":593,"linkText":400,"meta":594,"navigation":402,"path":595,"pinned":396,"publishTime":596,"seo":597,"stem":598,"tags":599,"verdict":609,"visual":419,"__hash__":610},"zeug_en\u002Fzeug\u002Fquooker-cube.md","Quooker Cube: expensive, unnecessary, and worth it every single day","Daily Driver",{"type":8,"value":427,"toc":582},[428,431,433,436,439,441,444,447,451,454,457,459,462,464,467,471,474,477,479,482,485,487,490,493,495,498,502,528,532,535,538,540,543,545,548,551,553,556,560,563,565,568,571,573,576,579],[11,429,430],{},"We drink a lot of water and a serious amount of tea.",[21,432],{},[11,434,435],{},"For both, there has basically always been a simple solution: a kettle for the tea, a crate of sparkling water from the basement.",[11,437,438],{},"It works.",[21,440],{},[11,442,443],{},"And yet there is now a tap in our kitchen that cost a multiple of what a normal tap costs.",[11,445,446],{},"And I would not give it back.",[52,448,450],{"id":449},"what-the-quooker-cube-actually-does","What the Quooker Cube actually does",[11,452,453],{},"The Quooker itself is, first of all, a tap that delivers boiling water at the push of a button.",[11,455,456],{},"Not hot. Boiling.",[21,458],{},[11,460,461],{},"The Cube is the unit that lives in the cabinet underneath. It turns the tap into a source of chilled still and chilled sparkling water as well.",[21,463],{},[11,465,466],{},"In practice: one tap, three things. Boiling water for tea, cold water for drinking and sparkling water, without ever carrying a crate again.",[52,468,470],{"id":469},"why-it-matters-so-much-day-to-day","Why it matters so much day to day",[11,472,473],{},"The real trick is not the technology.",[11,475,476],{},"It is the \"right now\".",[21,478],{},[11,480,481],{},"Tea is not a weekend event in our house, it runs all day. That used to mean: switch on the kettle, wait, forget about it, boil again.",[11,483,484],{},"Now I hold the cup under the tap and the tea is basically ready instantly.",[21,486],{},[11,488,489],{},"With water it is the same thing, cold. No hauling crates, no empty bottle at the worst possible moment, no deposit-bottle Tetris in the hallway.",[11,491,492],{},"Sparkling water is just there.",[21,494],{},[11,496,497],{},"This sounds like a luxury problem, and it is one. But it is one of those small things you use several times a day without ever thinking about it again.",[52,499,501],{"id":500},"what-i-like-about-it","What I like about it",[253,503,504,510,516,522],{},[256,505,506,509],{},[92,507,508],{},"Boiling water with no waiting."," Tea, coffee, quickly pouring over something. The kettle just gets in the way now.",[256,511,512,515],{},[92,513,514],{},"Sparkling water without crates."," No hauling, no deposit, no \"oh, the last one is empty\".",[256,517,518,521],{},[92,519,520],{},"One tap instead of three solutions."," Kettle, water crate and filter jug are suddenly redundant.",[256,523,524,527],{},[92,525,526],{},"It becomes normal immediately."," After a week you wonder how it ever worked before.",[52,529,531],{"id":530},"the-honest-downsides","The honest downsides",[11,533,534],{},"The price is steep. There is no way to make that sound reasonable.",[11,536,537],{},"A Quooker with a Cube costs a multiple of what you would purely functionally spend on hot and sparkling water.",[21,539],{},[11,541,542],{},"On top of that: this is built-in tech under the sink. Not a gadget you move somewhere else on a whim, but a small decision for the house.",[21,544],{},[11,546,547],{},"And it needs CO2 for the sparkling water.",[11,549,550],{},"The good news: you are not stuck with the expensive original bottles. Compatible CO2 bottles from third-party suppliers are a lot cheaper, and that makes the running cost bearable.",[21,552],{},[11,554,555],{},"Still, honestly: if you rarely drink tea and do not care about sparkling water, none of this makes sense.",[52,557,559],{"id":558},"would-i-recommend-it","Would I recommend it?",[11,561,562],{},"Yes. But with a clear if.",[21,564],{},[11,566,567],{},"If water and tea run all day in your home anyway, the Quooker Cube is one of the few expensive things that does not feel like waste, but like everyday life.",[11,569,570],{},"If, on the other hand, you rarely need hot water and are happy with tap water, stay away. Then it is simply a very nice, very expensive tap.",[21,572],{},[11,574,575],{},"For us it was an investment that made me swallow hard at first.",[11,577,578],{},"But boiling water for tea and cold sparkling water at the push of a button are worth it every single day.",[11,580,581],{},"And that is exactly the kind of good stuff this site is about.",{"title":383,"searchDepth":384,"depth":384,"links":583},[584,585,586,587,588],{"id":449,"depth":387,"text":450},{"id":469,"depth":387,"text":470},{"id":500,"depth":387,"text":501},{"id":530,"depth":387,"text":531},{"id":558,"depth":387,"text":559},"Gadget","2026-06-29","My experience living with the Quooker Cube: instant boiling water for tea, chilled sparkling water on tap and whether the steep price is actually worth it.","\u002Fimages\u002Farticles\u002Fquooker-cube-cover.webp","https:\u002F\u002Fwww.quooker.de\u002Freservoir\u002Fcube",{},"\u002Fzeug\u002Fquooker-cube","09:30",{"title":424,"description":591},"zeug\u002Fquooker-cube",[600,601,602,603,604,605,606,607,608],"quooker","quooker cube","kochendes wasser","sprudelwasser","wasserhahn","küche","tee","gadget","erfahrung","An absurdly expensive tap that never feels spectacular, and is worth it every single day for exactly that reason.","KaRZVbheXBtVYZsi4rZivTOViz7dHfkzEH9AtJahloA",{"id":612,"title":613,"badge":425,"body":614,"category":393,"date":907,"description":908,"draft":396,"extension":397,"image":909,"link":400,"linkText":400,"meta":910,"navigation":402,"path":911,"pinned":396,"publishTime":596,"seo":912,"stem":913,"tags":914,"verdict":922,"visual":419,"__hash__":923},"zeug_en\u002Fzeug\u002Fgmail-zu-kalender.md","Gmail to Calendar: The Only Email Automation I Actually Wanted",{"type":8,"value":615,"toc":898},[616,619,622,624,627,630,632,635,639,642,645,647,650,653,656,659,662,665,669,672,675,677,680,683,686,689,693,696,699,702,704,707,710,716,719,722,725,728,732,735,738,741,744,747,749,752,755,758,761,769,773,776,779,782,784,814,817,820,823,827,830,833,835,838,864,867,870,872,875,878,880,883,886,889,892,895],[11,617,618],{},"I do not want AI to answer my emails.",[11,620,621],{},"At least not automatically. I have enough respect for my inbox and for misunderstandings that the idea feels roughly as relaxing as giving a robot my house key and a bad mood.",[21,623],{},[11,625,626],{},"What I do want: appointments should stop disappearing inside emails.",[11,628,629],{},"Doctor appointments, contractor windows, delivery notices, club things, contract things, medical practice things. Somewhere in there is a date, usually a time, sometimes a location, and my brain says: I will do that in a minute.",[21,631],{},[11,633,634],{},"Spoiler: Usually not.",[52,636,638],{"id":637},"the-problem-is-not-email-the-problem-is-hiding","The problem is not email. The problem is hiding.",[11,640,641],{},"Email is fine for many things.",[11,643,644],{},"For appointments, it is weirdly bad.",[21,646],{},[11,648,649],{},"An appointment inside an email is not an appointment yet. It is a piece of text I have to read, understand, remember, transfer into the calendar, and ideally not copy incorrectly.",[11,651,652],{},"In a normal single-person household, that is already annoying.",[11,654,655],{},"In a family calendar, it quickly becomes absurd. If it is not in there, it basically does not exist. It may still be somewhere in Gmail, but Gmail does not get anyone to the concert on time.",[11,657,658],{},"That was the gap I wanted to close.",[11,660,661],{},"Not with a huge AI that organizes my life.",[11,663,664],{},"With a small automation that asks one question better than I do: does this email look like an appointment?",[52,666,668],{"id":667},"why-normal-email-notifications-do-not-help","Why normal email notifications do not help",[11,670,671],{},"The obvious solution would be: more notifications.",[11,673,674],{},"That is almost always the worst solution.",[21,676],{},[11,678,679],{},"I do not need another push notification telling me that something in an email might be important. I also do not need a filter that marks every semi-official email as important. Then everything is important, and my brain goes back to doing what it does best: ignoring it.",[11,681,682],{},"The problem is not that I get too few signals.",[11,684,685],{},"The problem is that the signals are too vague.",[11,687,688],{},"An email with an actual appointment should be treated differently from a shipping confirmation, a newsletter, or a polite reminder that some terms of service were updated somewhere. The latter may be legally fascinating. For my life, less so.",[52,690,692],{"id":691},"why-direct-calendar-automation-would-be-risky","Why direct calendar automation would be risky",[11,694,695],{},"The other obvious solution would be: AI writes appointments directly into the calendar.",[11,697,698],{},"Sounds efficient.",[11,700,701],{},"I hate it immediately.",[21,703],{},[11,705,706],{},"A calendar is not a notepad. Especially not a family calendar. If something is in there, other people plan around it. A wrong appointment is not just a small data error. In the worst case, it becomes an actual everyday problem.",[11,708,709],{},"Emails are also mean.",[11,711,712,713,95],{},"They do not simply say: ",[63,714,715],{},"Dentist, Tuesday, 10:30",[11,717,718],{},"They say: the original appointment is cancelled, the new suggestion would be Tuesday, Thursday would also work, please confirm by tomorrow. Or: the event is not at the location in the letterhead, but in room 2.14. Or: the date is just a deadline, not an appointment.",[11,720,721],{},"I do not want that in my calendar without checking it first.",[11,723,724],{},"Automation is good.",[11,726,727],{},"Blind automation is a very fast way to create very precise nonsense.",[52,729,731],{"id":730},"the-good-middle-ground-detect-summarize-ask","The good middle ground: detect, summarize, ask",[11,733,734],{},"That is why my Gmail-to-Calendar automation is deliberately boring.",[11,736,737],{},"It should not take over my inbox.",[11,739,740],{},"It should not reply.",[11,742,743],{},"It should not delete anything.",[11,745,746],{},"It should also not secretly add appointments.",[21,748],{},[11,750,751],{},"It checks whether an email probably contains an appointment. If it does, it pulls out the relevant pieces: occasion, date, time, location, and the context for why it thinks this could be a calendar entry.",[11,753,754],{},"Then it asks.",[11,756,757],{},"That last step is the important one.",[11,759,760],{},"Not because I enjoy adding another click. But because that click turns an automation into a tool. The AI detects. I decide.",[11,762,763,764,768],{},"That is the same kind of AI usefulness I like about the ",[33,765,767],{"href":766},"\u002Fen\u002Fzeug\u002Fhermes-agent","Hermes Agent",": no big future theatre, just one annoying little handoff made cleaner.",[52,770,772],{"id":771},"what-works-well-about-it","What works well about it",[11,774,775],{},"The magic is not that AI can operate a calendar.",[11,777,778],{},"A script can do that if it has to.",[11,780,781],{},"The useful part happens before that.",[21,783],{},[253,785,786,792,798,808],{},[256,787,788,791],{},[92,789,790],{},"Appointments become visible."," An email stops being just an email and becomes a concrete question: should this go into the family calendar?",[256,793,794,797],{},[92,795,796],{},"The context comes with it."," Not just date and time, but also occasion, location, and the short reason why the email might be relevant at all.",[256,799,800,803,804,807],{},[92,801,802],{},"I search less."," No later ",[63,805,806],{},"Where was that again?",", no frantic digging through Gmail right before we need to leave.",[256,809,810,813],{},[92,811,812],{},"The calendar stays clean."," Because nothing is added without approval, the family calendar remains a place for real decisions, not AI guesses.",[11,815,816],{},"All of that sounds small.",[11,818,819],{},"It is.",[11,821,822],{},"But those small handoffs are often exactly where everyday things get lost.",[52,824,826],{"id":825},"what-deliberately-stays-unautomated","What deliberately stays unautomated",[11,828,829],{},"I do not want AI acting as if it has authority.",[11,831,832],{},"Especially not with email.",[21,834],{},[11,836,837],{},"So a few things intentionally stay human:",[253,839,840,846,852,858],{},[256,841,842,845],{},[92,843,844],{},"No unchecked calendar entry."," The automation may suggest, but it does not decide on its own.",[256,847,848,851],{},[92,849,850],{},"No email replies."," Nothing gets confirmed, cancelled, or politely phrased in the wrong direction.",[256,853,854,857],{},[92,855,856],{},"No interpretation as truth."," If an email is ambiguous, it is ambiguous. Then I need to look at it.",[256,859,860,863],{},[92,861,862],{},"No productivity religion."," The goal is not to optimize every second of my life. The goal is to make one recurring small mistake happen less often.",[11,865,866],{},"And yes: with something like this, you have to think honestly about access and privacy.",[11,868,869],{},"An automation that reads emails is not harmless just because the word AI is involved. For me, the idea only works if the access is narrow enough, the purpose stays clear, and a calendar helper does not suddenly become an everything-reader.",[52,871,559],{"id":558},[11,873,874],{},"Yes.",[11,876,877],{},"But not as a big AI productivity promise.",[21,879],{},[11,881,882],{},"I would recommend it if appointments regularly land in your email and you then have to move them into a calendar manually. Especially if multiple people depend on that calendar and a forgotten entry does not stay your problem alone.",[11,884,885],{},"I would not recommend it if you already have a perfectly maintained calendar system, barely use email as an appointment source, or generally feel uneasy about automation with email access. That is legitimate. The gut gets a vote with these things.",[11,887,888],{},"For me, this automation is exactly right because it pauses at the decisive point.",[11,890,891],{},"It does not pretend AI is now my secretary.",[11,893,894],{},"It only says: this looks like an appointment. Do you want to take it over?",[11,896,897],{},"And sometimes the best automation is exactly the one that asks before it does something.",{"title":383,"searchDepth":384,"depth":384,"links":899},[900,901,902,903,904,905,906],{"id":637,"depth":387,"text":638},{"id":667,"depth":387,"text":668},{"id":691,"depth":387,"text":692},{"id":730,"depth":387,"text":731},{"id":771,"depth":387,"text":772},{"id":825,"depth":387,"text":826},{"id":558,"depth":387,"text":559},"2026-06-14","My experience with a small AI automation that detects appointments in Gmail, summarizes them, and asks before anything lands in the family calendar.","\u002Fimages\u002Farticles\u002Fgmail-zu-kalender-cover.webp",{},"\u002Fzeug\u002Fgmail-zu-kalender",{"title":613,"description":908},"zeug\u002Fgmail-zu-kalender",[915,916,917,918,919,920,921],"gmail to calendar","gmail automation","google calendar","ai automation","ai agent","family calendar","productivity","Not AI taking over my email. Just AI saying: this looks like an appointment.","HDYRM95kqcs65tW0IUT4gmiAY-3C4OkC-555CjJtTpc",{"id":925,"title":926,"badge":425,"body":927,"category":393,"date":1242,"description":1243,"draft":396,"extension":397,"image":1244,"link":1245,"linkText":400,"meta":1246,"navigation":402,"path":1247,"pinned":396,"publishTime":1248,"seo":1249,"stem":1250,"tags":1251,"verdict":1261,"visual":1262,"__hash__":1263},"zeug_en\u002Fzeug\u002Fvercel-typescript-hosting.md","Vercel: my favorite host for TypeScript projects",{"type":8,"value":928,"toc":1233},[929,932,935,938,940,943,945,952,954,957,961,964,967,969,972,980,983,985,988,992,995,998,1000,1003,1005,1008,1010,1013,1015,1018,1020,1023,1027,1030,1033,1036,1039,1042,1045,1048,1051,1054,1056,1059,1062,1065,1068,1071,1073,1111,1113,1116,1119,1121,1124,1127,1130,1133,1137,1140,1142,1145,1147,1150,1152,1155,1158,1160,1163,1165,1168,1170,1173,1175,1178,1181,1185,1188,1191,1193,1196,1199,1202,1206,1208,1211,1213,1216,1219,1222,1225,1227,1230],[11,930,931],{},"Hosting is actually one of those topics I do not like thinking about.",[11,933,934],{},"Not because it is unimportant.",[11,936,937],{},"More the opposite.",[21,939],{},[11,941,942],{},"When hosting is annoying, suddenly everything is annoying: deployments, previews, environment variables, domains, logs, builds, edge cases, rollbacks and that one tiny thing that of course refuses to just work on Friday at 4:47 pm.",[21,944],{},[11,946,947,948,951],{},"That is exactly why I like ",[92,949,950],{},"Vercel"," so much.",[21,953],{},[11,955,956],{},"For me, privately and at work, Vercel is the host where TypeScript projects move fastest from “runs locally” to “is online” without immediately turning into an infrastructure hobby.",[52,958,960],{"id":959},"why-typescript-fits-so-well","Why TypeScript fits so well",[11,962,963],{},"I build almost everything web-related in TypeScript somehow.",[11,965,966],{},"Nuxt, small internal tools, landing pages, APIs, admin surfaces, experiments, prototypes. Sometimes planned cleanly, sometimes honestly more like: I want to quickly see whether this idea holds.",[21,968],{},[11,970,971],{},"Vercel fits exactly that range.",[11,973,974,975,979],{},"Connect a Git repository, let it detect the framework, set environment variables, get preview deployments, put a domain on it, done. Of course in real projects it is never ",[976,977,978],"em",{},"only"," that simple. But the entry often feels like that.",[11,981,982],{},"And that is worth a lot.",[21,984],{},[11,986,987],{},"Especially with TypeScript projects, I do not want to first pet a server, explain build paths and meditate over deployment YAML. I want to push and see whether the app works.",[52,989,991],{"id":990},"preview-deployments-are-everyday-protection","Preview Deployments are everyday protection",[11,993,994],{},"The strongest Vercel feature for me is not the most spectacular one.",[11,996,997],{},"It is Preview Deployments.",[21,999],{},[11,1001,1002],{},"Every branch, every pull request, every change can exist as a real URL. Not as “imagine this were online”, but as an actually clickable thing.",[21,1004],{},[11,1006,1007],{},"At work, that is worth gold.",[21,1009],{},[11,1011,1012],{},"You can show features before they are merged. You can collect feedback on a real interface. You can test things without explaining the same local setup to every person involved.",[21,1014],{},[11,1016,1017],{},"Privately, it is just as pleasant.",[21,1019],{},[11,1021,1022],{},"Especially with projects like this site, I want to see how an article, an image or a layout actually looks when deployed. Not only locally. Not only in the good feeling. In the real hosting context.",[52,1024,1026],{"id":1025},"vercel-takes-the-boring-parts-off-my-plate","Vercel takes the boring parts off my plate",[11,1028,1029],{},"Vercel does many things I do not want to decide from scratch every time.",[11,1031,1032],{},"Builds.",[11,1034,1035],{},"Caching.",[11,1037,1038],{},"CDN.",[11,1040,1041],{},"Framework optimizations.",[11,1043,1044],{},"Serverless Functions.",[11,1046,1047],{},"Routing Middleware.",[11,1049,1050],{},"Environment Variables.",[11,1052,1053],{},"Domains and SSL.",[21,1055],{},[11,1057,1058],{},"Those are not small things.",[11,1060,1061],{},"But Vercel manages to make them feel small in everyday work.",[11,1063,1064],{},"When I deploy a Nuxt or Next app, I do not have to start from zero. Vercel recognizes a lot automatically and wires the usual pieces together so I can work first.",[11,1066,1067],{},"Of course you can argue about details. Of course there are limits, cost questions and cases where your own infrastructure or another provider is better.",[11,1069,1070],{},"But my default is still: if it is TypeScript on the web, I think of Vercel first.",[52,1072,501],{"id":500},[253,1074,1075,1081,1087,1093,1099,1105],{},[256,1076,1077,1080],{},[92,1078,1079],{},"Git push feels close to production."," Not ten intermediate steps, but a pretty direct path.",[256,1082,1083,1086],{},[92,1084,1085],{},"Preview URLs are incredibly practical."," For work, feedback, clients, colleagues and my own checks.",[256,1088,1089,1092],{},[92,1090,1091],{},"Framework support is strong."," Next is obvious, but Nuxt and other modern setups feel at home there too.",[256,1094,1095,1098],{},[92,1096,1097],{},"Serverless Functions are often enough."," Small APIs, webhooks, backend logic and integrations do not immediately need their own server project.",[256,1100,1101,1104],{},[92,1102,1103],{},"Domains and SSL are pleasantly unspectacular."," Exactly how it should be.",[256,1106,1107,1110],{},[92,1108,1109],{},"Logs and deployments are quickly reachable."," When something is broken, I do not want to first read an infrastructure treasure map.",[21,1112],{},[11,1114,1115],{},"And maybe that is exactly the point: Vercel does not feel like a separate hosting world.",[11,1117,1118],{},"It hangs directly on the way I already work.",[21,1120],{},[11,1122,1123],{},"Change code.",[11,1125,1126],{},"Push.",[11,1128,1129],{},"Look.",[11,1131,1132],{},"Improve.",[52,1134,1136],{"id":1135},"the-limits-are-still-real","The limits are still real",[11,1138,1139],{},"I do not want to romanticize Vercel.",[21,1141],{},[11,1143,1144],{},"If a project needs very specific infrastructure, long-running processes, unusual runtime requirements or full cost control on every level, then Vercel is not automatically the best answer.",[21,1146],{},[11,1148,1149],{},"Especially if you come from the “old world” of web development, where you have a virtual machine with a disk and a whole operating system available, you will stare stupidly at one or two places in the serverless world first.",[21,1151],{},[11,1153,1154],{},"Just upload a file in a form?",[11,1156,1157],{},"Just send an email?",[21,1159],{},[11,1161,1162],{},"Yes, technically these things all somehow work the “old way”, but not sensibly and certainly not well for long.",[21,1164],{},[11,1166,1167],{},"Then services like Vercel Blob\u002FS3, Postmark and so on quickly enter the picture, because they solve real problems you maybe did not even know you had before.",[21,1169],{},[11,1171,1172],{},"But that does not change the fact that Vercel hits exactly the right sweet spot for a lot of TypeScript web projects.",[21,1174],{},[11,1176,1177],{},"Not “everything is fine”.",[11,1179,1180],{},"More like: the normal things are solved normally.",[52,1182,1184],{"id":1183},"why-i-keep-landing-there","Why I keep landing there",[11,1186,1187],{},"I use Vercel privately because it gets my small projects from my head onto the web faster.",[11,1189,1190],{},"I use Vercel at work because Preview Deployments, clean integrations and fast iteration make real team life easier.",[21,1192],{},[11,1194,1195],{},"And I like Vercel because on many days it turns hosting back into what it should be for me: infrastructure in the background, not a side career.",[11,1197,1198],{},"When I work on an idea, I want to think about product, text, UI, data and users.",[11,1200,1201],{},"Not about the 14th YAML block.",[52,1203,1205],{"id":1204},"would-i-recommend-vercel","Would I recommend Vercel?",[11,1207,874],{},[11,1209,1210],{},"For TypeScript web projects, Vercel is my absolute default.",[21,1212],{},[11,1214,1215],{},"Not for every system.",[11,1217,1218],{},"Not for every architecture.",[11,1220,1221],{},"But for very many projects where deployment should be fast, clean and repeatable.",[11,1223,1224],{},"Vercel is good stuff for me because it gets deployment out of the way without feeling like a cheap shortcut.",[21,1226],{},[11,1228,1229],{},"It is a host that understands the way I work.",[11,1231,1232],{},"And that is surprisingly rare.",{"title":383,"searchDepth":384,"depth":384,"links":1234},[1235,1236,1237,1238,1239,1240,1241],{"id":959,"depth":387,"text":960},{"id":990,"depth":387,"text":991},{"id":1025,"depth":387,"text":1026},{"id":500,"depth":387,"text":501},{"id":1135,"depth":387,"text":1136},{"id":1183,"depth":387,"text":1184},{"id":1204,"depth":387,"text":1205},"2026-06-11","Why Vercel is the most pleasant host for TypeScript web projects, Nuxt, Next and fast deployments for me, both privately and at work.","\u002Fimages\u002Farticles\u002Fvercel-typescript-hosting-cover.webp","https:\u002F\u002Fvercel.com",{},"\u002Fzeug\u002Fvercel-typescript-hosting","09:15",{"title":926,"description":1243},"zeug\u002Fvercel-typescript-hosting",[1252,1253,1254,1255,1256,1257,1258,1259,1260],"vercel","hosting","typescript","next.js","nuxt","deployment","frontend","serverless","web development","Vercel is the host where I can focus on my project instead of all the surrounding infrastructure.","halftone","yvjvdxtN01OtNdLOQichdGf0sdCzk8jvPDWRMUgMMXw"]