Big Deal - Best Deals, Discounts, and Offers Online

Ultra Mobile by T-Mobile 4GB 12-Month Plan for $9/month + free shipping

<img src='https://d.dlnws.com/64599/1773131584-ultra-mobile-logo-7qek.webp?h=125&w=125' style='float: left;vertical-align: top;margin: 0 8px 8px 0'><div class="snippet summary" title="Ultra Mobile by T-Mobile's Back to School Savings sale discounts select multi-month plans for new customers. Save 15% on 6-month 4GB+ plans or 30% on 12-month 4GB+ plans, with the 4GB 12-Month Plan dropping to just $9.10 per month. All plans come with reliable 5G on T-Mobile, the nation's largest 5G network.  Offer ends August 31."><p>Ultra Mobile by T-Mobile's Back to School Savings sale discounts select multi-month plans for new customers. Save 15% on 6-month 4GB+ plans or 30% on 12-month 4GB+ plans, with the 4GB 12-Month Plan dropping to just $9.10 per month. All plans come with reliable 5G on T-Mobile, the nation's largest 5G network. Offer ends August 31. Buy Now at Ultra Mobile</p></div> <div class="snippet features" title="Unlimited talk to 90+ international destinations Unlimited talk and text in Mexico and Canada Free 5G access on the T-Mobile network Mobile hotspot included with all plans Supports eSIM and 3-in-1 SIM cards"> <div class="content-section-header">Features</div> <ul class="content-section-list"> <li>Unlimited talk to 90+ international destinations</li> <li>Unlimited talk and text in Mexico and Canada</li> <li>Free 5G access on the T-Mobile network</li> <li>Mobile hotspot included with all plans</li> <li>Supports eSIM and 3-in-1 SIM cards</li> </ul> </div>

Greenworks 40-Piece Impact Rated Driving Bit Set for $7 + free shipping w/ Prime

Amazon offers the Greenworks 40-Piece Impact Rated Driving Bit Set for $6.99, its best-ever price. It's also a $10 low. Shipping is free for Prime members. Buy Now at Amazon

American Tourister NXT 28" Hardside Expandable Spinner Suitcase for $45 + free shipping

eBay offers the American Tourister NXT 28" Hardside Expandable Spinner Suitcase for $44.99 via promo code "BULLSEYESAVE10". That's a $25 low. Shipping is free. Deal ends September 1st. Buy Now at eBay

Skibidi Toilet Mystery Surprise Toilet for $15 + free shipping

This Skibidi Toilet Mystery Surprise Toilet is $15.29 with promo code "BULLSEYESAVE10" for a $25 low. Shipping is free. Deal ends September 1st. Buy Now at eBay

Progress Lighting Avec Merrit 3-Light Chandelier for $29 + free shipping

eBay offers the Progress Lighting Avec Merrit 3-Light Chandelier for $29.44 via promo code "VIPAUGTAKE8". That's $90 under what Walmart charges. Shipping is free. Deal ends September 1st. Buy Now at eBay

Sabre Runner Pepper Gel Spray for $8 + free shipping w/ Prime

<img src='https://d.dlnws.com/64599/1788136486-81em18kktnl.jpg?h=125&w=125' style='float: left;vertical-align: top;margin: 0 8px 8px 0'><div class="snippet summary" title="Amazon offers the Sabre Runner Pepper Gel Spray for $8.49.  That's Amazon's best price of the past several years.  Shipping is free for Prime members."><p>Amazon offers the Sabre Runner Pepper Gel Spray for $8.49. That's Amazon's best price of the past several years. Shipping is free for Prime members. Buy Now at Amazon</p></div> <div class="snippet features" title="Adjustable 360° reflective hand strap for a secure grip 16-ft. spray range with up to 13 seconds of spray time Thumb trigger with twist-lock safety to prevent accidental discharge Holds 0.67 fl. oz. of pepper gel"> <div class="content-section-header">Features</div> <ul class="content-section-list"> <li>Adjustable 360° reflective hand strap for a secure grip</li> <li>16-ft. spray range with up to 13 seconds of spray time</li> <li>Thumb trigger with twist-lock safety to prevent accidental discharge</li> <li>Holds 0.67 fl. oz. of pepper gel</li> </ul> </div>

College Inn 32-oz. Chicken Broth for $2 + free shipping w/ Prime

Amazon offers the College Inn 32-oz. Chicken Broth for $1.63. That's Amazon's best price, and you'd pay around double this price locally. Shipping is free for Prime members. Buy Now at Amazon

H2OGO! Jump & Joy Bounce House w/ Slide and Blower for $40 + free shipping

Use promo code "VIPAUGTAKE8" to drop the H2OGO! Jump & Joy Bounce House to $40.48 for a $39 low. Shipping is free. Deal ends September 1st. Buy Now at eBay

S’well just discounted new travel sets, bottles and more

S'well is offering deals on its products throughout the month of December. The brand has released new travel container sets designed for organization on the go. These travel sets are compact, ...

El Pollo Loco Is Celebrating "12 Days of Pollo" with Delicious Deals

--This holiday season, El Pollo Loco is bringing back their wildly popular" 12 Days of Pollo"* celebration! Loco Rewards members can unlock a new holiday deal every day from December 1 through December 12 as El Pollo Loco kicks off the fourth annual 12 Days of Pollo. Dec. 1– BOGO Fire-Grilled Chicken Burrito Dec. 2– $10 Classic Tostada Combo Dec. 3– 20% Off a Purchase Dec. 4– $8 3-...

#ad 🔥Search my code[alh029037] NOW to unlock a AED400 coupon bundle on Temu! New App Users only#temu

ad Search my code [alh029037] NOW to unlock a AED400 coupon bundle on Temu! New App Users only. 1 per user. Coupons ...

This $20 Bill Is Actually a Clever Business Coupon

Credit: barberiacapuchino, jeronimo.casasola This clever marketer designed flyers that look like $20 bills, but the back reveals a ...

🚨 COUPON DU JOUR VALIDÉ ! 🚨 #betting #codepromo #coupondujour #footactu #football #gagnerauxparis

COUPON DU JOUR VALIDÉ ! La team, notre coupon 1XBET du jour a été VALIDÉ AVEC SUCCÈS ! ✓ Félicitations à ...

detroit clearance

OMG

Threshold Two Harbors 14" Patio Tabletop Fireplace for $33 + free shipping

eBay offers the Threshold Two Harbors 14" Patio Tabletop Fireplace for $33.29 via promo code "BULLSEYESAVE10". That's a $23 low. Shipping is free. Deal ends September 1st. Buy Now at eBay Features 15,000-BTU heat output includes lava rock

Show HN: BentoPDF, Hyper Compress and Kura

Hello. I developed an open source tool called BentoPDF. Its an open source PDF toolkit that runs in your browser. With the latest update, you can actually edit existing pdf text, images and objects right in your browser.<p>Live Website: <a href="https://www.bentopdf.com" rel="nofollow">https://www.bentopdf.com</a> Repository: <a href="https://github.com/alam00000/bentopdf" rel="nofollow">https://github.com/alam00000/bentopdf</a><p>Along with the latest update I would like to share with you Hyper Compress. Its a high fidelity, content preserving compression engine that preserves PDF conformance and surpasses all the other open source PDF compression tools.<p>It runs everywhere: CLI, Node SDK, C API, self hosted service, and also in browser via WebAssembly.<p>Live Website: <a href="https://hyper.bentopdf.com" rel="nofollow">https://hyper.bentopdf.com</a> Repository and benchmarks: <a href="https://github.com/alam00000/bentopdf-hyper-compress" rel="nofollow">https://github.com/alam00000/bentopdf-hyper-compress</a><p>Kura is a PDF standards, conversion and preflight engine.<p>It supports:<p>- All 11 PDF/A conformance levels: PDF/A-1a, PDF/A-1b, PDF/A-2a, PDF/A-2b, PDF/A-2u, PDF/A-3a, PDF/A-3b, PDF/A-3u, PDF/A-4, PDF/A-4e and PDF/A-4f<p>- Accessibility: PDF/UA-1 and PDF/UA-2<p>- Print production: PDF/X-1a, PDF/X-3, PDF/X-4, PDF/X-4p, PDF/X-5g, PDF/X-5n and PDF/X-5pg<p>- Engineering and variable data printing: PDF/E-1 and PDF/VT<p>- E-invoices: Factur-X, ZUGFeRD, XRechnung and Order-X<p>- 396 bundled print-preflight profiles<p>It has been tested against several standards suites, including the veraPDF corpus, Isartor, BFO, Ghent Output Suite 5.0, the PDF/UA Reference Suite and Cal Poly's PDF/VT suite.<p>Across 30,677 PDF conversions it had zero crashes and zero timeouts, with a 0.05 second median conversion time.<p>Like Hyper, it ships as a CLI, C library, npm package, Docker image and WebAssembly build.<p>Live Website: <a href="https://kura.bentopdf.com/" rel="nofollow">https://kura.bentopdf.com/</a> Repository and benchmarks: <a href="https://github.com/alam00000/bentopdf-kura" rel="nofollow">https://github.com/alam00000/bentopdf-kura</a><p>Both Kura and Hyper Compress are running the WASM build so none of your PDF is uploaded and everything runs in your browserr<p>If possible I would like you guys to try them out and give me feedback on how it worked for you.<p>Thank you.

Why the Information You Are Waiting for May Not Exist Yet?

That should be the ideal of everyone who demands certainty, and if you refuse to do it, because you “have enough of a certainty to know how things would go”, then don’t pretend you demand certainty, you only demanded reason to preemptively quit, because every prediction is based on a methodology, and the predictive power is relative to what the model quantifies over….I hope that albeit my obscure and at moments disconnected writing style, you understood what needs to be understood, that the inf

Show HN: Murmell – Collaborative cloud canvas for coding agents

And you can interact wich each other sessions, it's pretty useful instead of sharing a screen, copying a prompt, etc.The hardest part was building the cloud infrastructure, trying to make it scalable, and to manage the sessions to make the best experience possible: when the VM shuts down, it snapshots everything and restores everything when you come back, even the conversation with your terminal, it's stored with a system that i built like obsidian wich also make you save a little bit

Show HN: Train 300M/32-Layer Model in 1.5GB RAM on Base M1 Mac

Hello again! Since my last post about Ullis, the project has gone through several changes as I was searching for the right architecture. As it turned out, KAN and Hyena were quite resource-heavy, and I couldn't get anything viable out of them. Then I tried RWKV, specifically the latest RWKV-8 Heron version with 1-bit ROSA activation. This ultimately proved to be the most viable architecture of all. As an example: on my 2020 MacBook Pro M1 with 8GB RAM and a 68GB/s memory bandwidth, I managed to start training a model with ~272 million parameters, a 2048 context length, and 32 layers—all within a 1.5GB memory footprint. On my specific hardware, this is still a highly taxing task. Due to the low bandwidth, the total latency per step is around half a minute, and the throughput drops to about 70 tokens per second. To be clear right away: I had to keep ROSA SAM on the CPU because it proved to be more efficient for these tasks than the GPU. Here are the logs as proof:<p>cargo run --release -- train \ --data data/ullis_dataset.jsonl \ --run runs/hard \ --config train_config.json \ --steps 20000 \ --learning-rate 0.005 \ --checkpoint-every 500 \ --bpe-train-mib 150<p>ullis: token stream compacted to u16 (156 MiB) ullis: compiling Metal shaders ullis: Metal ready in 0.2s ullis: Metal train: LN/QKV/CMix/head on GPU; ROSA SAM on CPU (~603979776 bytes idx+y+out/step) ullis: starting loop after 379.0s of setup ullis: clipped SGD lr=0.005 rosa_grad=StopGradBits (QKV frozen; window-mean CE on FP16, token-sum STE on BinaryConnect; |w0|=0.01) step 1/20000 loss=7.9256 ema=7.9256 p10=5.218 p50=8.193 p90=10.197 unigram=5.958 unique=597 n=1389 flips=0/0/0 (head/cmix/o) embed_grms=6.40e-5 scale_grms=8.09e-3 cmix_vrms=0.013 |w|=0.010 dw=4.37e-6 bias_rms=2.033 resid=3.21e-8 rss=1643MiB 27881ms 73 tok/s ullis: step 1 phases embed=8ms fwd_ln=370ms fwd_rosa=7220ms fwd_cmix=5897ms head=1070ms bwd_ln=625ms bwd_cmix=9754ms bwd_rosa=2678ms embed_sgd=57ms step 2/20000 loss=7.0537 ema=7.8384 p10=1.670 p50=7.518 p90=9.969 unigram=5.564 unique=590 n=1520 flips=0/0/0 (head/cmix/o) embed_grms=6.46e-5 scale_grms=1.06e-2 cmix_vrms=0.013 |w|=0.010 dw=4.68e-6 bias_rms=2.033 resid=5.67e-8 rss=1628MiB 27997ms 73 tok/s ullis: step 2 phases embed=8ms fwd_ln=314ms fwd_rosa=5717ms fwd_cmix=4738ms head=1231ms bwd_ln=786ms bwd_cmix=11755ms bwd_rosa=3180ms embed_sgd=60ms<p>With scaled-down settings, training runs with pretty satisfactory performance. Here is an example: since a multi-stage training pipeline with pre-training is not implemented yet, I trained it on a large Claude Opus distillation dataset. After 5500 steps, here is the result:<p>cargo run --release -- generate \ --checkpoint runs/ullis_b1_32m/checkpoint.safetensors \ --prompt "2+2" \ --temperature 0.4 \ --top-p 0.8<p><pre><code> Finished `release` profile [optimized] target(s) in 0.60s Running `target/release/ullis generate --checkpoint runs/ullis_b1_32m/checkpoint.safetensors --prompt 2+2 --temperature 0.4 --top-p 0.8`</code></pre> - *No a single = = = = = = i < (r: $=3_x(d: $d = \frac{[b_t + 2>[b_t + 1 + 3-c)$<p>- The correct answer is a list of the number of the first n > 0<p>As you can see, the model is capable of learning, but since I cannot afford to keep my main work laptop running training tasks 24/7, I had to stop at this relatively modest result. There is, of course, a possibility that some training algorithms might have implementation bugs, but this is the current outcome. I hope you find this project interesting. Honestly, pulling off a project like this entirely on your own is quite tough, especially considering I'm developing it alongside an LLM assistant (which constantly tries to break things). I'm not deeply experienced in this field yet, so I'd be incredibly glad to find testers, people with domain expertise, or anyone willing to share any kind of feedback. Thank you!<p>Link: <a href="https://github.com/Vladislav-Kalinkin/ullis" rel="nofollow">https://github.com/Vladislav-Kalinkin/ullis</a>

Show HN: My startup-idea scanner scored 500 ideas; the best got 6.3/10

I built a tool (<a href="https://1mil.app/hn" rel="nofollow">https://1mil.app/hn</a>) to help me test my intuition about various weird business ideas that were on my mind. Initially, I wanted to create a database with 1 million business ideas (hence the name - "1 million opportunities"). But turned out to be a dumb idea already. So instead I built this tool that returns only 10 ideas. It's all based on your input, market + your background.<p>The first release scored on demand only and it gave false confidence. So I rebuilt the the scoring around "winnability": can a solo founder with a specific background "win" against what already exists?<p>Heres what still bugs me: since winnability went live in June, it has scored 500+ ideas, and none reached 7/10. The best was 6.3 and the median is 1.3. That includes every scan I ran for myself.<p>My preliminary conclusion: most business ideas are improbable to win.<p>Well, go ahead and try the scanner, and let me know what you think. A free scan without an account is included. Each scan costs Vertex tokens, which is why the tool isn't free.