Exact 12-hour window: **2026-09-24T18:01:10.686231+00:00 to 2026-09-25T06:01:10.686231+00:00** (20:01:10.686231 on 24 September to 08:01:10.686231 on 25 September, Europe/Warsaw, CEST). Feed timestamps below are UTC.
This is a **nonexhaustive one-page sample: one latest page of up to 40 posts per monitored account, without pagination**. The supplied topic sample contains 111 posts; relevant, interpretable tech posts are reproduced below, while unrelated posts, context-free reactions and unsupported personal accusations are omitted. Retweets and truncated excerpts are not independent corroboration. All claims, figures, release names and availability statements come from these posts and have not been independently verified; linked media and external articles were not inspected.
Local inference optimization dominates this sample: @sudoingX reports faster Bonsai 2 decoding on consumer GPUs, while @TheAhmadOsman and @ivanfioravanti disagree over how much early Apple hardware comparisons reveal before kernel tuning. Agent workflows are another focal point, with posts describing Claude cloud sessions, automatic Amp runners and code-derived architecture maps. @kunchenguid offers a personal model-routing framework rather than a controlled comparison or official release announcement. Security and business signals—including an Ubuntu update-cadence claim and a Higgsfield revenue figure—deserve follow-up but remain unverified social-source assertions.
- **Consumer-GPU inference:** @sudoingX says a restored MTP head and kernel changes raise Bonsai 2 27B performance on an RTX 3060 from 26 to 50 tokens/s at fresh context, with 18.3 tokens/s at 77k context. The author also claims 10,756 downloads; these are self-reported measurements and adoption figures, not independent tests. → link
- **Early Mac versus Spark comparison:** @TheAhmadOsman reports two DGX Sparks achieving 2.41× faster prefill than a Mac Studio M5 Ultra on DeepSeek V4 Flash, with the Mac 6.4% ahead in generation and time-to-first-token figures of 1.3s versus 2.9s. The post itself cautions that Mac kernels are not optimized; configuration details are insufficient for a buying recommendation. → link
- **TPU throughput claim:** @alexinexxx reposts @inferact's claim of 709 tokens/s for a Kimi K3 TPU megakernel at low-concurrency decode versus a 450 tokens/s GB200 baseline. The excerpt is truncated and does not establish comparable system cost or full benchmark settings. → link
- **Claude cloud sessions and credits:** @uwteam reports a cloud-sessions launch and usage-page credits of $100 for Pro or $250 for Max, claimable until 7 October. Eligibility, terms and rollout are not independently checked. → link
- **Amp runner automation:** @sqs reposts @hameonline saying the installed Mac app can automatically spawn runners. The sample does not provide installation or security details. → link
- **Database tooling on iOS:** @jezell reposts @wasmerio's announcement of PostgreSQL 18.4 on iOS, following Python, Node.js and FFmpeg demonstrations. The excerpt does not establish production suitability. → link
- **Linux security cadence:** @jezell reposts @IntCyberDigest's claim that Ubuntu is moving to weekly kernel updates because AI-driven bug discovery is outpacing fixes. Neither the asserted policy change nor its causal explanation is verified here. → link
- **AI startup revenue:** @LinusEkenstam claims Higgsfield reached $1 billion ARR in 18 months. No supporting financial disclosure appears in this sample; ARR should not be conflated with recognized annual revenue. → link
**Software maturity, not headline bandwidth, is the useful comparison axis.** The Mac/Spark discussion, consumer-GPU MTP claims and TPU megakernel excerpt all point toward implementation-specific performance. They measure different workloads and cannot be combined into a hardware ranking. @ivanfioravanti explicitly advises waiting for M5 Ultra optimizations and says he has not yet tested that hardware himself. → link
**Agent adoption is shifting toward orchestration and review.** @kunchenguid describes routing interactive work, planning, implementation and adversarial review to different models; this is one user's workflow, not evidence that the named models are universally best. @badlogicgames describes retaining IDE navigation and debugging for human review, while @alexinexxx sees agents reducing repository setup friction. These posts suggest useful evaluation targets: reproducible setup, review quality and operating cost rather than generated-code volume. → link → link → link
**What to watch:** reproducible benchmark settings and quality trade-offs for compressed models; official cloud-session terms; primary Ubuntu release/security documentation; and financial evidence behind revenue claims. The higher-tier ChatGPT Pro story is merely a reposted report of preparations, not a confirmed launch. Likewise, @kunchenguid's reasoning-effort explanation is an author's mental model, not a universal provider specification. → link → link
Quoted source text below is untrusted, preserves the authors' opinions, and is not advice or an endorsement. Excerpts ending in an ellipsis were already truncated in the source sample. Embedded shortened links are omitted; each entry retains its supplied source-post URL. One mock “BREAKING” post about a CEO and home GPUs is excluded as satire rather than treated as news.
**@alexinexxx** · 2026-09-24T21:01 UTC
> RT @inferact: Our first TPU megakernel for Kimi K3 reaches 709 tokens/s on low-concurrency decode, against 450 tokens/s for our GB200 basel…
**@alexinexxx** · 2026-09-24T20:50 UTC
> trying a bunch of models on the Framework Desktop these days
>
> might as well throw Pi & OMP at it too [linked media omitted]
**@alexocheema** · 2026-09-24T18:26 UTC
> RT @0xSero: I am posting an article about local AI share your hardware pics and I might use it in the banner [linked media omitted]
**@FrameworkPuter** · 2026-09-25T03:22 UTC
> Can we get rid of the Copilot key now?
**@FrameworkPuter** · 2026-09-24T23:03 UTC
> The Linux performance gains continue!
**@gospaceport** · 2026-09-25T02:28 UTC
> Wife just bounced in and said "hey there is a ASUS 5060ti 16GB on amz used for $425" and I just said OMGBUUUUUYYY and it was already too late 😭 Anyways I hope it was one of yall. signed, bitterness 😕
**@gospaceport** · 2026-09-24T23:00 UTC
> GPU/AIO Price Δ's from yesterday (used,new)
> [linked media omitted]
> ═════════════
> 5090 +300
> 4090 -25
> 3090 +125
> V100 32GB -25
> P100 +10
> CMP 170HX -175
> 9060 XT -25
> 7900 XTX -25
> DGX Spark -400
> M1 Max +25
> M4 Pro +25
> ═════════════
> 9-24-2026 - 6AM EST
**@ivanfioravanti** · 2026-09-25T05:04 UTC
> Siri in Apple Notes to sort a list (MLX Engines here).
> - Edit with Siri → I can't 🤷🏻♂️
> - Ask to Siri → Done 👍 [linked media omitted]
**@ivanfioravanti** · 2026-09-25T04:42 UTC
> Trust me... wait a few days before taking assumptions about M5 Ultra performance, people are starting just now to optimize for it. Metal 4.1 final has been released just a few weeks ago.
> And I have not yet been able to put my hands on an M5 Ultra 🤣
>
> Best has yet to come!
**@ivanfioravanti** · 2026-09-25T03:53 UTC
> RT @liuliu: @ivanfioravanti Yeah, like I commented elsewhere, @localcodeapp already gets to ~900 tok/s prefill on M5 Max. No reason M5 Ultr…
**@ivanfioravanti** · 2026-09-25T03:52 UTC
> I think now Inference Engines are mature on Qwen 3.8 Flash Next (and therefore ready for Qwen 4).
>
> Time for a royal rumble! I will run an LLM Context Benchmark with prose and coding and post results!
>
> Which engines (MLX based) should I include in the run?
**@ivanfioravanti** · 2026-09-25T03:11 UTC
> Wait for M5U optimizations to be in place before drawing any conclusion.
> Clearly decode is not optimized yet.
**@MilksandMatcha** · 2026-09-24T19:03 UTC
> Cerebras puts power delivery half a millimeter from the wafer. I asked Angela Yueng, SVP of Product, why that tiny distance matters so much.
>
> In part two of our AMA, we got into the hardest parts of engineering CS4, how @cerebras is using AI agents to monitor its systems, and whether there’s still room to push the wafer faster.
>
> 0:27 Why half a millimeter matters for AI speed
> 2:11 AI agents monitoring the hardware that runs AI
> 3:06 The hardest part of building CS4
> 4:16 Have we hit the wafer’s power ceiling
> 5:26 How many countries does it take to build one CS4
**@MilksandMatcha** · 2026-09-24T18:56 UTC
> RT @cerebras: CS4 uses the same wafer as CS3. So where does the extra speed come from?
>
> You asked on Twitter and Discord. Join Angela Young…
**@sudoingX** · 2026-09-25T05:08 UTC
> meanwhile there are people arguing about whether one used rtx 3090 is worth it.
>
> the men at the top are hoarding compute like it's the only thing that will matter. buy a gpu. buy the cheapest one you can find. just get on the board.
**@sudoingX** · 2026-09-25T04:15 UTC
> don't sleep on this model gamers, you already have the vram it needs, and you can taste the freedom of local ai on the card sitting in your pc right now.
>
> it runs a 27b model at 50 tok/s fresh on gpus like rtx 3060, way faster than you can read, with no account and nothing leaving your machine. and it doesn't just chat, it does agent work, you hand it a task and it opens files, writes the code, runs it, checks it and keeps going while you're away from the desk.
>
> this chart is the part people ask about most, context. with the mtp head on it starts at 50 tok/s and still does 18.3 tok/s with 77k tokens already in the window, faster than you read. the full 262k window fits on a 12gb card with the head off, and the shipped mtp script runs 163k, that's a whole codebase in one prompt. it slows as the window fills, and i'd rather show you the whole curve than crop it.
>
> the card you bought for games sits idle most of the day, give it a job. i packaged all of it for you, the model, a serve script for every vram tier and a prebuilt bundle, so you extract it and run.
>
> it got 10,756 downloads in 4 days already. 3060, 3070, 4060, anything with 8gb and up, pull it and tell me what your card does:
>
> [linked media omitted]
**@sudoingX** · 2026-09-25T03:20 UTC
> thank you x for the balloons 🎈 and thank you to everyone who pulled bonsai 2 this week, 10,756 downloads is the best birthday gift i could ask for [linked media omitted]
**@sudoingX** · 2026-09-25T02:30 UTC
> how many computers do you own that are currently doing absolutely nothing?
**@sudoingX** · 2026-09-25T00:20 UTC
> holy shit! my bonsai2 mtp gguf now has 10,756 downloads in under 4 days, 6,780 of them since last night, for a model i packaged for gaming gpus like rtx 3060, 3070, 3080s and 40s. i am still processing this.
>
> if you're new here, this is bonsai 2 27b, qwen 3.8 27b dense compressed to ternary by prismml. it shipped without its mtp head, so i grafted the head back on and shipped a faster kernel for the gaming cards, that took it from 26 tok/s to 50 tok/s on a 3060.
>
> this week it built octopus invaders, a multi file space shooter, on a 12gb card, 5 hours, 8 files, every line written by the model, and the downloads went vertical right after. i'll leave the build link below.
>
> the most common gpu on steam is a 3060 and the whole point was that card, the one already sitting in your pc, running a model this size at home with nothing leaving the machine.
>
> model card and the scripts for your card are below, pull it and tell me what your gpu does.
**@sudoingX** · 2026-09-24T21:39 UTC
> 6 days ago i told the 3060 crowd to wake up. here's what the same card does now:
>
> > 26 tok/s fresh then, 50 tok/s now with my kernel and the mtp head
> > 13 tok/s at 77k context then, 18.3 tok/s now
> > a gpu monitor in 77 minutes then, a whole multi file game overnight now
> > not even released then, 10,756 downloads now
> > 8gb cards run it too, 64k context
> > my kernel got approved upstream by prismml
>
> same five year old gpu card, same bonsai 2 weights, everything that changed was software. the most owned gpu on steam just keeps getting better, pull it tonight if you have one.
**@TheAhmadOsman** · 2026-09-25T04:13 UTC
> Get used to a Local AI where everyday your tokens per second gets better and better
**@TheAhmadOsman** · 2026-09-25T02:37 UTC
> The optimizations in MLX are horrid, DHH is right Apple doesn’t have a hardware problem they have a software problem
**@TheAhmadOsman** · 2026-09-25T01:17 UTC
> Just a reminder that I said this almost a year ago, and now we have something amazing, free, and made by the people for the people at @OsmanticAI
>
> If I say it’ll be done, you can mark me on it
>
> Opensource AI will win and Local AI will be the default.
**@TheAhmadOsman** · 2026-09-24T22:47 UTC
> Most notable thing
>
> M5 Ultra with 1.2 TB/s bandwidth is only 6.4% faster in token generation than the DGX Spark (273 GB/s)
>
> M5 Ultra has ~5x the Bandwidth of the DGX Spark
**@TheAhmadOsman** · 2026-09-24T22:28 UTC
> Mac Studio M5 Ultra vs 2x DGX Spark
> for DeepSeek V4 Flash
>
> - DGX Sparks: 2.41x faster on prefill (compute-bound)
> - Mac Studio M5 Ultra: 6.4% faster on generation (bandwidth-bound)
>
> Time to first token
> - Spark 1.3s
> - Mac 2.9s
>
> Of course, the Mac M5 doesn't have optimized kernels [linked media omitted]
**@alexinexxx** · 2026-09-24T23:45 UTC
> feels like coding agents quietly expanded the open source ecosystem by making “figure out how to run this repo” an agent task
**@badlogicgames** · 2026-09-24T21:35 UTC
> RT @groma_md: Hello world! 👋 Meet groma.md.
>
> Your architecture as a live C4 map: scanned from your code, curated by your coding agent, and…
**@badlogicgames** · 2026-09-24T18:41 UTC
> something @giuseppegurgone taught me at the first claude code anonymous, many hundred years ago.
>
> i don't use my bespoke review tool anymore. i inspect stuff in VS Code, with code nav/debugging/whatever i need to grok the code, then leave BOZOs for the agent. [linked media omitted]
**@FinansowyUmysl** · 2026-09-24T18:48 UTC
> RT @FinansowyUmysl: Powiem Wam, że Opus 5.5 wymiata w porównaniu do Fable.
>
> I nie chodzi o jakość - moim zdaniem porównywalna.
>
> Chodzi o pr…
**@gospaceport** · 2026-09-25T01:20 UTC
> Hermes Agent is the agent for ppl that like to update agents [linked media omitted]
**@ivanfioravanti** · 2026-09-25T05:57 UTC
> Time to test Opus 5.5!
**@jezell** · 2026-09-24T21:55 UTC
> RT @wasmerio: Yesterday we announced Python, Node.js and FFmpeg running on your iPhone.
>
> Today: a real PostgreSQL 18.4 server. On iOS. And…
**@jezell** · 2026-09-24T21:02 UTC
> RT @btibor91: OpenAI is preparing a new higher-tier ChatGPT Pro plan [linked media omitted]
**@jezell** · 2026-09-24T20:34 UTC
> RT @huacnlee: GPUI has long lacked a complete, systematic set of documentation. GPUI Kit has the same gap—something the community has been…
**@jezell** · 2026-09-24T18:56 UTC
> RT @IntCyberDigest: ‼️ Ubuntu is moving to weekly kernel updates because AI is finding Linux bugs faster than Canonical can ship the fixes.…
**@KingBootoshi** · 2026-09-25T01:30 UTC
> how to become a better agentic engineer as models improve:
>
> > try a prompt you don't think is gonna work
>
> > watch it work
>
> > expand your perception of what is possible
>
> rinse and repeat every new model release
>
> at this point, the only limit is psychological HAHAHAHA
**@KingBootoshi** · 2026-09-25T01:00 UTC
> you WERE a software engineer
>
> NOW you are an architect of reality
**@KingBootoshi** · 2026-09-25T00:56 UTC
> i mean really? we have thought to reality compilers and people are complaining?
>
> do they have NO THOUGHTS? AT ALL? NOT AN OUNCE OF CREATIVITY IN THEIR BONES?
>
> ITS BAFFLING
**@KingBootoshi** · 2026-09-25T00:55 UTC
> it seems that ai is revealing most people actually hate their careers and never had passion for it in the first place
>
> if you truly had passion and curiosity for your field ai should feel like God's gift dawned amongst men
>
> but instead most are threatened because it was something they were 'good' at, and now they feel worthless
>
> i hate to break it to you bud, if that's your mindset then you definitely are worthless now
**@KingBootoshi** · 2026-09-25T00:50 UTC
> software engineers 'quitting' because ai makes them feel useless are hilarious to me
>
> they just weren't worthy enough to wield the thought-to-reality compiler
>
> do they have no thoughts? no grand visions of builds? ambitions?
>
> clankas take away the reason to write code manually and they feel like a dud?
>
> HOLY SKILL ISSUE
**@kunchenguid** · 2026-09-25T03:29 UTC
> how to set the right reasoning effort level
>
> i get asked this a lot, and i learned that because the underlying mechanism is not obvious, a lot of people don’t have the right mental model to think about how to use the knob
>
> so a quick crash course here:
>
> reasoning effort level in mainstream LLMs today means a “ceiling”, not a “floor”!
>
> setting a high effort level does NOT mean every prompt you send will use a lot of thinking tokens. you can try this very easily - launch your agent in an empty directory, use high reasoning, and say “hi” - unless your harness is totally messed up, the agent will respond almost instantly
>
> how does that work? the most typical implementation is that in post-training, the models receive a penalty for how many thinking tokens they use. how strong the penalty is depends on the thinking effort level that’s set
>
> at low reasoning, the penalty is very strong, so any long thinking traces get punished into oblivion
>
> at high reasoning, the penalty is weak so longer thinking traces don’t get punished too much, and are allowed to survive more often as long as they get good results
>
> the most important thing to learn there is - thinking tokens is always a penalty, not a reward. there’s never a rule that says “at high reasoning i’ll punish you if you think too little” - this is because more thinking tokens always means higher cost which is always a bad thing and should not be rewarded
>
> so with that in mind, you should think of reasoning effort as “how much are you allowed to think”, not “how much do i want you to think”
>
> setting it at low means “for this task you are not allowed to think much”. you would want this when 1) the use case requires low latency; or 2) you already know the task does not require additional reasoning
>
> setting it at high means “you are allowed to think more when needed”. the model will still try to think as little as possible according to what its training data teaches it. you want this when 1) the use case allows slower response; and 2) you are not sure how much thinking is needed for this task
>
> for some models, “max” is special because it forces thinking mode to be on, which forbids “zero thinking”. i don’t have a full inventory for which models do this vs not. and i generally avoid this mode because i don’t think it’s a good idea to say “every request has to have thinking tokens no matter what”
>
> practically speaking, i mostly only use two modes:
>
> - medium, when i already know the task is well defined
>
> for example, implementing a spec planned by a highly intelligent model. i avoid low because 1) i rarely require low latency; 2) even when implementing well defined tasks sometimes the intermediate context is still ambiguous
>
> this is also my default for firstmate because highly ambiguous problems are typically handled by crewmates, and i do want firstmate to be fast
>
> - xhigh, whenever the task is not well defined yet. this is most often used for planning and investigative work. i do this to give enough room for the model to decide how much thinking it needs
>
> hope this is helpful!
**@kunchenguid** · 2026-09-25T00:14 UTC
> alright! as the dust settles around this insane week of model releases, i've stabilized around a new model line-up so sharing here for reference
>
> this time my approach is a bit more structured. i've bucketed various LLMs into a few categories:
>
> 1. interactive orchestrators
>
> these are models that i use as firstmate (and second mates), because they are fast, efficient, pleasant to talk to, intelligent enough to understand my intent, and have good enough judgment to steer the crew around it
>
> for me, this bucket is opus 5.5 and grok 4.7
>
> viable budget alternatives when my subscription quota runs out: muse spark 1.3, deepseek v4 flash
>
> 2. premium intelligence
>
> these are models that i only use for highly ambiguous or creative tasks that i decide to truly need the extra intelligence and justifies the cost
>
> i also use this bucket to handle escalations - when crewmates started arguing with each other, when a review-loop started spiraling out of control, when a simple change somehow ended up with a giant PR - i call these models to untangle the mess
>
> for me, this is currently gpt 6 astra and fable 5.1 (or opus 5.5 when fable quota is tight)
>
> 3. planners
>
> these models are the ones i trust as default for planning new features and investigating complex bugs. they would produce a spec that get implemented by a cheaper model
>
> i almost exclusively use opus 5.5 for this right now because of its incredible ROI, whenever i don't need the premium intelligence
>
> 4. implementers
>
> these are efficient workhorses that when given a well defined spec they can produce solid implementation
>
> i use opus 5.5, gpt 6 sol, and grok 4.7 for this right now, and again the budget options: muse spark 1.3 and deepseek v4 flash (i'm sure there are many other viable alternatives as well - i just haven't got enough time to try them out)
>
> among these models i currently find sol to be the best adversarial code reviewer (i haven't tried astra for this, as it's a bit too expensive to run at such high volume)
>
> 5. trivial fixers
>
> these are the models i use for extremely trivial changes like a one-liner fix or config change. they really don't need much intelligence because often times what needs to be changed was already defined
>
> for these i use gpt 6 luna and again the budget options if quota is tight
>
> the way i actually make use of the categorization is that i told these routing preferences to firstmate, and it can then help me route the right task to the right model (which is now made very efficient because of Jev)
>
> one last thing i'll point out is you can see how versatile opus 5.5 is in my line up - it can do pretty much everything! this is the first model that spanned across almost every bucket in my setup, which is quite a massive advantage because it means if i have to choose only one subscription it would have to be anthropic at the moment
**@kunchenguid** · 2026-09-24T21:04 UTC
> RT @deimantasnork: Hey captain @kunchenguid,
>
> look what I built for Firstmate on @fabienpenso’s Herdr-gpui edition 👀
>
> Captain’s Deck: live…
**@kunchenguid** · 2026-09-24T18:20 UTC
> this is the first week ever since fable release that my claude weekly overall quota is running out faster than fable quota
>
> opus 5.5 seriously changed the game [linked media omitted]
**@levelsio** · 2026-09-24T23:04 UTC
> RT @IterIntellectus: holy shit
> i asked claude to make a video on western civiization [linked media omitted]
**@LinusEkenstam** · 2026-09-24T20:19 UTC
> yo fam, this Opus 5.5 LEGO is wild
>
> LEGO should legit create a bricks API. I know we’ve got BrickLink.
>
> But LEGO could open up a whole new massive business vertical here.
>
> Imagine being able to hook this up to an official LEGO API, where you can order whatever you can dream up.
>
> LEGO then packs your specific built to order box, and charge per brick with a small premium.
>
> Absolutely absurd what could come out of this.
>
> @LEGO_Group should pay attention.
**@LinusEkenstam** · 2026-09-24T19:53 UTC
> Designers now have a lot of new surfaces to consider:
>
> - iPhone Duo
> - Meta VR Glasses
> - Agentic use
> - Audio
> - Tamagotchis
>
> To orchestrate a brand across all existing surfaces + everything new. All while the tools, tasks, workflows are in hyper flux is the most exciting times ever to touch design.
>
> I recently visited California, and I had not booked a place to stay. I gave my agent an address I wanted to be within 2miles of. I tasked it to look up all sorts of sites, booking, hotels, lastminute, airbnb, kayak.
>
> The second request was to make sure it kept within budget, and that whatever it decided to put infront of me was actually available.
>
> The deliverable, a little website, with all the things it found, with some simple filters, distance from adress, price (all inclusive, tax, cleaning fee etc)
>
> I did not trust the agent to book me a place just yet. But trusted it enough to build me that website.
>
> That solidified my confidence in that everything will become a data layer. Because I don’t really care for certain utilities, but the counter point I’m also pretty confident about is that brand will matter more than ever.
>
> In a world where everyone can get an optimized result for their current need. People will long for to be surprised and delighted. Thus world class brands will thrive at the same time we will have pure magic experience at our fingertips.
>
> Learning how to compose and maintain a strong and clean brand across all surfaces will be paramount.
>
> What do you think, will brands still matter in 10 years?
**@louszbd** · 2026-09-25T03:25 UTC
> One last /workflow dogfood before mid autumn festival 🥮 Looking promosing so far, off to celebrate now. [linked media omitted]
**@MengTo** · 2026-09-25T05:10 UTC
> RT @MengTo: This looks straight out of Blender, but it's all code in a single HTML file at about 186 KB.
>
> Astra can create these ultra-real…
**@nummanali** · 2026-09-24T23:15 UTC
> Perfect opportunity to spend time with Muse models
**@ollama** · 2026-09-24T20:06 UTC
> RT @CrusoeAI: Closed model, open model, or train your own?
>
> 📅 Wed, Oct 7
> 🕐 5 PM
> 📍 San Francisco
>
> At our next AI Talk, Crusoe’s @ErwanM94…
**@RayFernando1337** · 2026-09-25T03:41 UTC
> RT @beyang: Great if you’re doing iOS or Mac app dev and need to get xcodebuild to do something
**@RayFernando1337** · 2026-09-25T00:25 UTC
> Removed 60,984 lines of test slop :) [linked media omitted]
**@sqs** · 2026-09-25T03:10 UTC
> Kids love Amp, too [linked media omitted]
**@sqs** · 2026-09-25T00:15 UTC
> RT @hameonline: Launched just now in amp - if you have the mac app installed, it can automatically spawn runners for you!
>
> No more pesky te…
**@sqs** · 2026-09-24T23:13 UTC
> RT @thorstenball: Time to share your Amp runners
>
> [linked media omitted] [linked media omitted]
**@steipete** · 2026-09-24T21:12 UTC
> RT @pejmanjohn: The Muse Charm launch is reminding me that I built this @openclaw tamagotchi for Apple Watch back in February but never lau…
**@steipete** · 2026-09-24T19:00 UTC
> RT @aesofaces: @steipete predicted the end of plan mode a loooooong time and now it’s officially here
**@swyx** · 2026-09-25T06:00 UTC
> RT @latentspacepod: This is tracking to be the #1 podcast we've released in all of 2026.
>
> We asked Jev's creator @CompleteSkeptic to explai…
**@TheAhmadOsman** · 2026-09-24T21:33 UTC
> RT @OsmanticAI: Still a week left to show up and participate in our Opensource Contributors program for September 💪❤️
>
> [linked media omitted]…
**@TheAhmadOsman** · 2026-09-24T21:05 UTC
> There’s a difference between having a Kofi link to support your Opensource contributions and begging btw
>
> We even have an Opensource Contributors program at @OsmanticAI where we spend thousands of dollars a month to support our Opensource Contributors ❤️
**@TheAhmadOsman** · 2026-09-24T19:02 UTC
> Wait, why is Zach not monetized? This guy actually posts real alpha, gives free GPUs credits to people doing interesting work / doesn’t beg for money, and actually does his homework before he talks so it’s not nonsense or stolen
>
> Go follow this guy and give him some views [linked media omitted]
**@uwteam** · 2026-09-24T18:42 UTC
> Anthropic znów rozdaje prezenty. W ustawieniach Claude w zakładce "usage" czeka prezent: $100 (Pro) lub $250 (Max). Czas na odbiór jest do 7 października.
>
> Bonus z okazji odpalonej dziś funkcji "cloud sessions" (coś jak cloud agents w Cursorze). Warto doczytać, co to jest i jak to działa.
**@jack** · 2026-09-24T21:17 UTC
> RT @ClaraShik: No soft forks, no bridges, no operators, just Bitcoin.
>
> We’ve put a lot of work into thinking carefully about the security…
**@jack** · 2026-09-24T21:14 UTC
> RT @allocinitxyz: Shielded Bitcoin: Private Transfers on Bitcoin L1 [linked media omitted]
**@jack** · 2026-09-24T21:07 UTC
> RT @BlockIR: Square has introduced a new integration with Apple Business in Australia, Canada, France, Ireland, Japan, Spain, the UK, and t…
**@jack** · 2026-09-24T21:05 UTC
> RT @Square: From new ways for customers to find you, to fewer steps during service. A preview of 8 new features that start rolling out next…
**@jack** · 2026-09-24T21:00 UTC
> lightning on x402
**@jack** · 2026-09-24T21:00 UTC
> RT @bitcoinpolicy: Freedom Tech DC: Open Intelligence with @Jack
>
> - Open source is a boon for global innovation
> - AI will reshape company s…
**@jack** · 2026-09-24T20:56 UTC
> RT @wesbillman: Buzz Desktop v0.5.25 🐝
>
> Browse canvas history, see who changed what, and restore an earlier version without losing the edit…
**@jack** · 2026-09-24T20:56 UTC
> RT @CashApp: Bitcoin bonus is live in Cash App.
>
> Opt in to earn weekly bitcoin* rewards when you hold bitcoin in Cash App. There’s no locku…
**@jezell** · 2026-09-24T20:28 UTC
> There is no point in making these if they don't do launch parties in Irvine Spectrum like the good ole days with @DustinBrowder and the SC team. Collectors edition without signatures from all the devs is not a true Blizzard collector's edition.
> [linked media omitted]
**@levelsio** · 2026-09-24T23:31 UTC
> 📡 Kinda anecdotal but I now have 2 friends with @Starlink (and so 3 if you include me)
>
> I loaned my Starlink Mini to @mick__net after his 🇵🇹 Portuguese ISP @NOS was (and still is!) down for 2+ months, while he lives 500 meters away from the NOS headquarters in Lisbon 😂 Nobody can fix it, customer support is useless and even cancelling his plan is difficult
>
> Anyway Starlink Mini got him his internet back immediately and saved his ass cause he had to work, and now he ordered his own and gave mine back, which I then used for traveling in Greece, and he's a Starlink convert
>
> Now today my other friend @cso_notes also got one in 🇷🇴 Romania!
>
> Adoption in Europe always comes a bit later than in US, but I think Starlink is starting to get a significant presence now
>
> Many people who I tell about Starlink are still skeptical about satellite internet, they always talk about latency issues, and they think latency is like 1 second but it's more like 10ms (1/100th of a second) which is the same or better than regular fiber internet
>
> I've never had so LITTLE issues with internet as since I had Starlink, it's never been down, it always just works and customer support is via chat and works really well
>
> I absolutely HATE having to call a phone number and be on hold for 30min or more and then wasting most of my day on stuff like this, which par for the course with Portuguese ISPs (and well every Portuguese business)
>
> So having Starlink is a godsend for me, it just works! 🙏
**@LinusEkenstam** · 2026-09-24T22:02 UTC
> $1BN in ARR in 18months 🤯🤯🤯
>
> So many people are throwing shade on Higgsfield, but call me crazy, hitting $1bn in ARR after 18months is world class.
>
> Clearly show that growth, is in fact the most important factor if you want to grow
>
> Who would have guessed 😂
>
> *NOT SPONSORED
**@swyx** · 2026-09-25T05:49 UTC
> In Jan this year I called my content strategy shot: "Scaling without Slop".
>
> It's finally starting to work. It took us 3 years to reach our first 100k on youtube. It only took 1.2 months for the next 100k. Similar other metrics on AEO/SEO/subscriber traction and have a lot of New Media ideas that I'm excited to pursue.
>
> officially giving notice of the next phase of Latent Space, AINews, and what the rest of swyx inc has been cooking below
**@TrungTPhan** · 2026-09-25T00:33 UTC
> RT @dee_bosa: Tomorrow: @satyanadella
>
> On YouTube, X and @YahooFinance. Daily show launches end of October. [linked media omitted]
**@TrungTPhan** · 2026-09-24T23:17 UTC
> Same team behind FoodNeverComes (lets you "experience" ordering food without having to pay or eat) also created a "Dopamine Site" called TripNeverLeaves.
>
> Some amazing taglines for fake trip-booking site:
>
> "The anticipation IS the holiday"
> "All the escape, none of the stress." [linked media omitted]
**@TrungTPhan** · 2026-09-24T20:53 UTC
> Tim Cook seeing Zuck ship high-fidelity Meta VR Glasses that weigh 1/6th of the Apple Vision Pro (100g vs. 600g) at 1/3rd the price ($3,699 vs. $1,299) and with a virtual keyboard that you can actually use on hard surfaces with normal typing motion instead of pinching the air. [linked media omitted]
**@TrungTPhan** · 2026-09-24T18:05 UTC
> RT @colossusjeremy: Besides being the greatest diarist of the tech age, @benthompson is an awesome interviewer and great hang. We talked ab…
**@uwteam** · 2026-09-25T05:57 UTC
> Moja przygoda z demonetyzacją na YouTube skończyła się happy endem.
>
> Odezwał się do mnie @DenizRymkiewicz z firmy Google i obiecał pomóc w tej sprawie - no i pomógł.
>
> Czas na nowo ułożyć sobie życie za moje dochody - 24zł miesięcznie 😎
>
> Tak poważniej mówiąc, to chodziło jedynie o to, aby mój kanał nie był oznaczony jako kategoria low-quality, czy stolen-content. Jednak okazało się, że ja to naprawdę ja, a moje treści jednak są moje :)
>
> Istnieje pewna niezerowa szansa, że zależało mi na odblokowaniu kanału, ponieważ planuję powrócić do nagrywania na tej platformie, ale to mogą być też jakieś plotki ;)
**@uwteam** · 2026-09-24T20:00 UTC
> RT @uwteam: Od wielu lat rozdaję darmowe serwery VPS - takie naprawdę malutkie.
> 256 MB RAM i 3 GB dysku to parametry śmieszne jak na dzisie…