Register and share your invite link to earn from video plays and referrals.

Search results for 140キロコース
140キロコース community
One keyword maps to one global community path.
Create community
People
Not Found
Tweets including 140キロコース
Eurozone shares +0.2% (-1.4% wk) US shares -0.4% (+0.1% wk) as strong payrolls boost case for Fed rate hike, Trump call to cut not helping! US 10 yr yld +1bp to 4.78% Oil +0.2% to $91.5 Gold -1% to $4430 Iron ore +1% to $100.3 Bitcoin $79.7k ASX futures -0.01% $A 0.72 w $US +0.3%
Show more
swift-transformers 1.4.0 has just been released 🚀 Focus: tokenization correctness, performance, and alignment to transformers. PRs from 15 different contributors, 10 of them new. Thank you! 🙏
Show more
It's only been a day since 0.7.1, but oMLX 0.7.2 is out with changes worth shipping now. This release brings faster DeepSeek V4.1 on every oQ build, Lightning MTP in distributed inference, and Qwen3.8-Flash-Next GGUFs on Splash 1.4.0. Faster DeepSeek V4.1 on oQ2e to oQ4e - oQ3e with Lightning MTP + CED: decode 49.4 -> 68.2 tok/s (+38%), prefill 700 -> 949 tok/s (+36%). Every quant from oQ2e to oQ4e gets faster, with similar prefill gains. - Prefill holds around 900 tok/s all the way to a 128K prompt. - Measured on M3 Ultra 512GB. I'll run more tests on M5 Ultra 256GB. Lightning MTP in Distributed Inference - DeepSeek V4.1 and GLM-5.3 run Lightning MTP across Macs with both PP and TP. - With TP on two M3 Ultras, both models decode more than 50% faster, even over a TCP ring. I'll share exact numbers after more tests over RDMA. - A pipeline stall is fixed, so DeepSeek V4.1 across two Macs went from about 1 tok/s to about 28 tok/s. More Models on Splash - Qwen3.8-Flash-Next GGUFs run on Splash 1.4.0, and oMLX offers the matching DFlash2 draft. - Split GGUFs download and show up as one model, and mlx-community OptiQ builds now start. Other Improvements - Better prompt cache reuse in Claude Code sessions, on both the built-in engines and Splash. - Reasoning models answer Claude Code's auto-mode classifier before it times out. - DFlash works with ParoQuant Qwen models. - SpecPrefill no longer disturbs other requests decoding next to it, and Gemma 3 gets the right positions in sparse prefill. - Keyboard access to the web UI menus and better contrast in the dark theme. Thanks to everyone who sent PRs and tested 0.7.1 so quickly!
Show more
🇬🇧 England vs Norway 🇳🇴 – All Historical Football Matches 1937: Norway 0-6 England 1938: England 4-0 Norway 1949: Norway 1-4 England 1966: Norway 1-6 England 1980: England 4-0 Norway 1981: Norway 2-1 England 1992: England 1-1 Norway 1993: Norway 2-0 England 1994: England 0-0 Norway 1995: Norway 0-0 England 2012: Norway 0-1 England 2014: England 1-0 Norway England wins: 7 Draws: 3 Norway wins: 2
Show more
CHIP STOCKS SLIDE The Philadelphia Semiconductor Index fell 3.6% in early trading. $MU -6.7% $ARM -4.6% $INTC -4.0% $AMD -3.5% $TSM -3.4% $MRVL -2.7% $NVDA -2.3% $AVGO -1.4%
libghostty has always been fast, but it's also been getting significantly faster. Here are IO throughput measurements comparing Ghostty 1.3.1, 1.4.0, and main as of this morning. We'll be backporting most of these to a 1.4.1 release soon. Here are some highlights of the changes since 1.4: - Ghostty 1.4 had a fast-path for printing runs of plain text. But the fast-path only applied to limited cases. We now support mixed width characters and hyperlinks in the fast path. - Our input path always widened the pty data to codepoints (8-bit to 32-bit) as part of UTF-8 decode, which also incurred a memory copy. We now have a SIMD fast path to detect pure-ASCII and we process the run direct from the read buffer (no copy, narrower data). - New memory pages use most recent two pages actual utilization to pre-size themselves accordingly. This sped up styles, hyperlinks etc. - When a VM page runs out of capacity for a certain type of data (e.g. hyperlinks, styles), we now project the usage so far to guess what you'll need, then shrink available rows to avoid a realloc. Before we'd naively double. This caused a classic "laddering" problem. - We recognized "erase" being a slow path for us, and it's a common sequence, so we use another bit in row headers to track blank (we had space). - Styles are now processed once per SGR. Previously, a sequence like "ESC[1;38;2;r;g;bm" would apply each style as it came in (1, then 38, then 2, etc.). We now apply all attributes at once and write to the cursor atomically. - Styles are now packed into 128-bit values and use aarch64 and x86_64-specific assembly for minimizing instructions to use it. - On x86_64, we found that LLVM was lowering some row resets (hot path) to a 4-byte store followed by an 8-byte load which couldn't be store-forwarded. This caused an almost 10% (!!!) slowdown on scroll operations only on x86_64. The fix was to change our reset to build the full header in a register then do a single memmove per row. Net results are what you see in the graph, but written out main to 1.4: ASCII 1.5x, heavy styling 1.5x, styled Unicode 1.8x, random UTF-8 1.4x, random OSC 8 2.7x, mixed styled text with links 4.8x, cursor/erase 16x, hyperlink-dense text 5.6x.
Show more
✨ Yamaga Seika: Boundless Imagination in Textile Art From rockets and constellations to Tokyo Tower, discover a world of imagination woven in thread. Celebrating the artist’s 140th birth anniversary, this first retrospective in 40 years explores the vibrant creativity of a pioneer of textile art. 🧵 📍 Venue: Kyoto City KYOCERA Museum of Art 🗓️Dates & Hours(UTC+9): September 19 – December 20 , 2026, 10:00–18:00(Last admission is at 17:30) Closed on Mondays (open if Monday is a public holiday) 🔗 Click to get your tickets and discover Yamaga Seika’s extraordinary woven world! Let every thread reveal a new wonder. ✨ ✨ 生誕140年記念 染織家 山鹿清華─宙翔ぶイマジネーション ロケットや星座、東京タワーまで。糸で織りなす、想像力あふれる世界へ。生誕140年を記念する40年ぶりの回顧展で、染織芸術のパイオニアが生み出した、鮮やかな創造の世界をたどります。🧵 📍 会場:京都市京セラ美術館 本館 南回廊1階 🗓️日時(UTC+9):2026年9月19日(土)~12月20日(日)10:00-18:00(最終入場は17:30まで)休館日:月曜日(祝日の場合は開館) 🔗 からチケットを購入して、山鹿清華が織りなす驚きの世界を体感してください! 一本の糸から、新たな驚きに出会うひとときを。✨ #YamagaSeika# #TextileArt# #山鹿清華#
Show more
During Tesla Cybercab launch week in Austin, I took a total of 12 paid Cybercab rides and kept every receipt. TOTAL: • 12 rides • 29.7 miles traveled • $108.34 total spent • $3.65 average cost per mile • 2 hours and 19 minutes riding around Austin in a Cybercab • Compared to Uber and Waymo, every Cybercab ride was still ~30-50% cheaper EACH OF MY CYBERCAB RIDES: 1/ Sept 3 — 7:23 PM 3.9 miles · 16 min · $8.70 ≈ $2.23/mile 2/ Sept 3 — 7:53 PM 0.5 miles · 6 min · $4.65 ≈ $9.30/mile 3/ Sept 4 — 6:52 AM 2.5 miles · 12 min · $6.99 ≈ $2.80/mile 4/ Sept. 4 — 7:13 AM 2.4 miles · 11 min · $6.84 ≈ $2.85/mile 5/ Sept. 4 — 7:39 AM 1.8 miles · 11 min · $6.16 ≈ $3.42/mile 6/ Sept. 4 — 8:03 AM 1.4 miles · 8 min · $5.69 ≈ $4.06/mile 7/ Sept. 4 — 8:20 AM 3.3 miles · 15 min · $7.90 ≈ $2.39/mile 8/ Sept. 4 — 10:58 AM 4.0 miles · 19 min · $8.78 ≈ $2.20/mile 9/ Sept. 4 — 11:58 AM 2.6 miles · 8 min · $7.08 ≈ $2.72/mile 10/ Sept. 4 — 2:17 PM 2.8 miles · 11 min · $13.71 ≈ $4.90/mile 11/ Sept. 5 — 12:57 PM 2.4 miles · 10 min · $16.18 ≈ $6.74/mile 12/ Sept. 5 — 1:36 PM 2.1 miles · 12 min · $15.66 ≈ $7.46/mile One of the biggest things I learned from actually using Cybercab is that you cannot judge the economics from just one ride. Pricing is dynamic. Some of my longer rides came in cheaper around $2.20–$2.40 per mile. Some of the shorter or higher-demand rides were much more expensive. Sometimes, the Cybercab was even more expensive than the Model Y (Unsupervised) bc the demand was so high. Some of the rides also became more expensive as Cybercabs became bookable on September 4 to the general public. However, it’s hard for me to look at these receipts and not wonder how Uber and Lyft can compete if Tesla can scale Cybercab the way it wants to. Uber and Lyft were built around connecting riders with human drivers. That means every ride has to support a much bigger cost stack... like 1/ A human has to be paid. 2/ That person needs a car. 3/ That car has to be financed or purchased. 4/ It needs insurance. 5/ Maintenance. 6/ Fuel or charging. 7/ And then Uber or Lyft still needs its cut. Cybercab attacks the biggest cost in that entire system which is the driver... there is no driver waiting to be paid at the end of the ride. And Tesla is supplying the software. Tesla can build the car. Build the battery. Build the autonomy system. Operate the network. Control the app. Control the charging. And the team designed the Cybercab from Day 1 around being a Robotaxi instead of taking a normal consumer car and trying to make the economics work. This is a big deal. Bc if Tesla can manufacture these at huge scale, keep them on the road for most of the day, and keep bringing the cost per mile down, then Uber and Lyft are going to be competing with the company that built the vehicle, the driver, and the network ALL AT ONCE. And this was during launch week... with just ~45 Cybercabs, before this vehicle is anywhere close to mature scale. If Tesla can push that cost down further as the fleet grows to thousands and millions, while Uber and Lyft still have to pay a human being to sit behind the wheel… I really don't see how these companies can survive... I think the days of Uber and Lyft are limited.
Show more
🔶 PS5 & PC Shared update details! 🔸As a token of appreciation for all the love and support from our players, we've prepared 6 newly arranged tracks based on existing BGM. Visit 'The Last Gulp' in Xion to receive 'A Thank You Gift From Enya' -The Song of the Sirens (Arrange) -Star Descent (Arrange) -Arisa (Arrange) -Democrawler (Arrange) -Everglow (Arrange) -Hypertube (Arrange) 🔸Fixed an issue in Orbital Elevator - 'CARGO LIFT 121' where resting during the elevator dialogue could occasionally cause the dialogue to restart or the elevator to freeze. 🔸Fixed an issue in Boss Challenge where Scarlet would disappear during a phase transition cutscene if Tachy Mode was activated. 🔸Fixed an issue in the NIKKE collaboration mini-game where using 'Burst Skill △' on specific monsters would make them immune to Eve’s ranged attacks, preventing progress. 🔸Fixed an issue where the reeling icon appeared gray after switching the fishing method from 'Reel In Controls' to 'Timing Controls' input. 🔶 PC // 1.4.0 New update details. 🔸Fixed an issue where, on PC, using a controller with 'Swap Confirm/Cancel Buttons' enabled would cause the UI to display incorrectly after completing a quest. 🔸Fixed an issue during the 'Oblivion' side-quest where the flooded water would disappear before reaching the sewage control device' step, intermittently preventing progression 🔸Improved key input handling so that Left/Right variants (L/R) of modifier keys (Ctrl, Alt, Shift, Cmd) are now recognized as the same input. 🔸Fixed an issue where dodging became unavailable if the player attempted to dodge while shooting after using 'Burst Skill X' in the NIKKE mini-game. 🔸Fixed an issue where a black box would appear on screen when playing on a Samsung Odyssey 3D monitor. 🔸Various other bug fixes. #StellarBlade# #PS5# #PC# #Update#
Show more
🧩 Qwen3.8-Flash-Next: A 6B-Active Preview of the Qwen4 Architecture On the same night Zhipu's GLM-5.3-Flash took over the timeline, @Alibaba_Qwen open-sourced Qwen3.8-Flash-Next — explicitly positioned as a preview of the Qwen4 architecture. Zhihu contributor Kitt在进化 argues that for people who actually run models locally, this is the more practical release of the night. The local-deployment groups he is in are, in his words, on fire. His take: the model is called 3.8, but the architecture is really Qwen4 in preview — and it borrows the best ideas from across the field. 1️⃣ Smaller, cheaper, and realistic for local deployment Flash-Next has 125B total parameters with only 6B active — far smaller than the 300B-class GLM-5.3-Flash. The API is priced at ¥1 input, ¥3 output, and ¥0.1 per cached million tokens, roughly matching DeepSeek-V4-Flash's off-peak rates. His comparison, per 1M tokens in RMB — input / output / cache hit: 🔹 Qwen3.8-Flash-Next: 1.0 / 3.0 / 0.1 🔹 GLM-5.3-Flash: 0.4 / 1.4 / 0.115 (limited-time promo rate) 🔹 DeepSeek-V4-Flash (off-peak): 1.5 / 4.5 / 0.05 2️⃣ A Qwen4 preview wearing a Qwen3.8 name The author points out this is the same play Zhipu made: GLM-5.3-Flash's architecture is also completely different from GLM-5.3. Shipping the new architecture as open weights early is deliberate pathfinding. Inference frameworks like vLLM and SGLang, plus the quantization toolchain, all need lead time to adapt before Qwen4 proper arrives. 3️⃣ An architecture that borrows from everyone The author reads Flash-Next as a synthesis of the field's best recent ideas. 🔹 Attention: Qwen's in-house QSA, built to balance throughput and speed on long context. 🔹 Knowledge: it absorbs DeepSeek's Engram line of work, packing prior knowledge into 51B of N-gram side parameters — high knowledge density at minimal compute cost. 🔹 Training: the Muon optimizer, popularized by Kimi, scheduled together with AdamW. 4️⃣ Half the active parameters, still overtaking At 6B active, Flash-Next is nearly half the size of DeepSeek-V4-Flash — yet Qwen's reported benchmarks show it overtaking that model on multiple coding and agent leaderboards. The author says he is not worried about real-world experience: the recent Qwen3.8-27B already proved itself in daily use, and this sits on the same foundation. 🔗 Full Reading: #Qwen# #Alibaba# #Qwen4# #OpenWeights# #LLM# #AIInference# #MoE#
Show more