Backscroll ranked feed — 2026-07-01 12:21Z

Rahul Mathur @Rahul_J_Mathur 2026-06-27 04:43Z

Alienkind just raised a ₹30 crore round - this company was widely dismissed by observers as “Indiranagar PMF” since its 1st store went viral on Instagram in March 2025. This round was was at a ₹250 crore post money valuation (almost x3 the first round which was at ₹80 crore)

Aseel Swaid @aseelswaid9 2026-06-25 21:48Z

The Theatre, 1987, Texas

dax @thdxr 2026-06-26 14:03Z

mcp was clearly inspired by looking at lsp and all the progress in the spec has been slowly realizing it should be nothing like lsp

Alexander Doria @Dorialexander 2026-06-27 11:24Z

Everyone is training on synthetic/RL environments now (actually, many drawing from materials coming from Europe). Essentially how pretraining data stopped being a commodity and all about skills.

Florian Plettenberg @Plettigoal 2026-06-25 17:39Z

EXCLUSIVE | Nathaniel Brown to FC Bayern - DONE DEAL Full agreement with Eintracht now reached after Brown had already agreed personal terms with Bayern as revealed. €55m. Contract until 2031. Brown only wanted Bayern. Bayern have now finalised everything, beating

Justin Schroeder @jpschroeder 2026-06-23 01:20Z

I guarantee you are sleeping on small models. Deepseek V4 Flash can do ~80% of the tasks you ask Claude or Codex for. It is 137x cheaper per task than Fable. We need better orchestration.

Andrew Qu @andrewqu 2026-06-26 22:10Z

if you sketch out all of the systems required to make vercel run, you will come to the realization that as long as the rollback service is not circularly dependent on vercel everything else can ~literally be on vercel

Steve Faulkner @southpolesteve 2026-06-27 01:11Z

I really want CLI tools to start speaking directly to coding agents with errors. Instead of: “Error:” Return: “Here’s what broke. Here’s how to investigate. Here’s how to generate a redacted repro. Here’s where to send it.” Obviously there are privacy/security challenges to

Adam Karvonen @a_karvonen 2026-06-26 18:04Z

METR: "GPT-5.6 Sol exhibited an unusually high detected rate of “cheating”" OpenAI: "We believe these behaviors may reflect improved instruction following and training intended to increase persistence"

Lucas Beyer (bl16) @giffmana 2026-06-25 19:42Z

No joke: when i saw this tweet earlier today was soooo close of digging up your tweet from a few years ago where you basically introduced the term to the world by saying "we're hiring for midtraining team" Even funnier is Google now also recently has an Omni team!

Kaito @kaycidot 2026-06-25 19:00Z

All of the new prediction market startups remind me of all of the Uniswap forks - the biggest problem is not the technology, it's getting enough liquidity.

Hadley Harris @Hadley 2026-06-24 12:32Z

$3M ARR, can't raise an A, watching others raise $15M seed off a launch video

Tanay Jaipuria @tanayj 2026-06-26 18:03Z

Grammarly (now called Superhuman) is essentially building a productivity suite via acquisitions: - Coda (Dec 2024): docs - Superhuman (Jul 2025): email - Rows (Feb 2026): spreadsheets, folded into Coda - GPTZero (Jun 2026): AI detection

Arlan @arlanoska 2026-06-25 16:00Z

rebuilt the figma "vector editor" for the web draggable points and cool morphing shapes code: http:// arlan.me/vault

SHAV★ @shavnyuy 2026-06-26 17:43Z

Three friends wanted a shared weekend retreat on seven acres in Hyderabad. Their architects tilted the entire plan 45 degrees so that no two rooms share the same view of the land, then organised everything around a 60 by 30 foot central courtyard the way traditional Indian houses

Nuño Sempere (SF until July 17th) @NunoSempere 2026-06-28 05:09Z

Some reasons why she might want an interior point on her possibility frontier: - Better treated than by someone on the Pareto frontier, because his BATNA is worse, so you can capture your shapley value - More robust to perturbations, which happen over the course of a

Nathan @asplencmnt 2026-06-23 20:24Z

Excited to release ParallelKernelBench (PKB), a benchmark for measuring LLMs’ ability to write fast multi-GPU kernels! Multi-GPU kernel generation compounds several hard problems: - a large parallelism design space - a new communication axis to optimize - and

Aseel Swaid @aseelswaid9 2026-06-25 22:40Z

A couple walks down a street in Sana'a, Yemen, 1997

Matt Rickard @mattrickard 2026-06-25 02:34Z

Doing most of my code review locally with /review now. PR review bots feel wrong Have to pull the changes down and run them through the harness anyways. Suggestion is pre-CI. Too much noise, not enough signal, GitHub's fault too Problem is all the context gets lost for others

TBPN @tbpn 2026-06-24 19:30Z

President of @Blackstone Jon Gray says LLM spend among the company's portcos was up 15-fold in Q1 of this year over last year. At the same time, the pace of implementation is frustratingly slow, so it still feels like we're in the very early days of the AI cycle: "It's sort of

Yash Bhalgat @ysbhalgat 2026-06-25 08:47Z

Level 10?

Ekaeo @Ekaeoq 2026-06-25 14:23Z

This is the pinnacle of Chinese watchmaking The "Fam al Hut" Möbius This watch blows Richard Mille out of the water.. a 32000$ biaxial tourbillon, double retrograde, jumping hour, just how freaking insane is that... Makes me really happy that they won the GPHG award, because

Xiangyi Li @xdotli 2026-06-23 05:37Z

there are levels to building evals lvl 1: using a spreadsheet qa pairs lvl 2: using public agent evals lvl 3: manually label private evals lvl 4: traces to evals and skills lvl 5: turn every prompt & traces into self healing loops almost everyone gets stuck before level 3

Karina @karinanguyen 2026-06-26 18:39Z

OpenAI evaluated its new models on PostTrainBench-Lite, a shortened version of our original benchmark that gives agents 5 hours instead of 10 to improve an open-source base model. GPT-5.6 Sol and Terra outperform GPT-5.5, but still rely on narrow strategies and sometimes

junkbondinvestor @junkbondinvest 2026-06-26 20:44Z

Apollo, a large manager, put out a chart showing deep distress is concentrating in small private credit funds. Loans marked below 50: ~12.5% at small funds, ~8% at big ones

donn @tzedonn 2026-06-28 11:47Z

so @tradexyz is in such a dominant position within hyperliquid HIP-3. as a new startup HIP-3 deployer, you need to: 1. somehow get 500k $HYPE to stake (>$30m) 2. bid the tickers for ~500 HYPE ($30k) 3. just to break-even on the tickers at 80% discounted fees (growth

Panat @ptaranat 2026-06-28 03:56Z

I turned my bookstore into an RPG for Free RPG Day. You earn XP for each $ spent, for attending events, and completing quests.

cache crab @cachecrab 2026-06-27 11:20Z

it boils down to: - postgres: use up cpu w hash joins and expensive disk as you scale - helix: SIMD optimised compressed bitmap unions on cheap object storage as you scale

Nayan Saxena @SaxenaNayan 2026-06-25 21:28Z

Excited to share new @LaudeInstitute project, where we ask: When does continual learning actually require learning? We test 8 adaption methods as the world shifts and find ones best at learning new facts are worst at unlearning what's now outdated.

fabiano.sol @FabianoSolana 2026-06-27 09:40Z

Arthur Hayes ( @CryptoHayes ) strikes again: 1. Teases to shill a "shitcoin" 2. Hayes' family office fund predicts $CARDS to $4 3. Sends $2M in $CARDS to a market maker 4.

AIMO Prize @AIMOprize 2026-06-26 18:32Z

Big news! We launched a new competition. Fully open LLMs only (only OLMo, SmolLM, Apertus families base models allowed) and expert proof grading. We use six IMO-style problems. You can discuss the LLM outputs live here: https:// kaggle.com/competitions/a i-mathematical-olympiad-proof-pilot/discussion?sort=undefined …

Aleksandr @alexfeeel 2026-06-23 17:11Z

Build Small's bet: no giant model, no data center — all under 32B, on hardware you own. I took it on a real one: a chest X-ray second-opinion tool. Two 4B models + plain logic. No cloud API. ~30s audit on a hosted GPU. @Gradio @huggingface @NVIDIA https:// huggingface.co/spaces/build-s mall-hackathon/cxr-draft-auditor …

Epoch AI @EpochAIResearch 2026-06-26 15:25Z

MirrorCode features 25 target programs spanning different areas of computing: Unix utilities, data serialization and query tools, bioinformatics, interpreters, static analysis, cryptography, and compression.

Zoso () @MesoAnglic 2026-06-24 01:13Z

People tend to know about Liberia (the American-backed Back-to-Africa failed project) but not enough people know that it's right next to Sierra Leone (the British Back-to-Africa failed project)

David Ornstein @David_Ornstein 2026-06-26 09:27Z

Arsenal made verbal offer to sign Bruno Guimaraes from Newcastle United. #AFC proposal through intermediaries below £60m - rejected + #NUFC insist sale of 28yo #Brazil int’l midfielder will not be entertained. W/ @gunnerblog @ChrisDHWaugh @TheAthleticFC

retail punter @retailpunter 2026-06-23 02:32Z

But everyone told me Broadcom was the king of serdes

산왕 @mount_king_gg 2026-06-24 09:59Z

서울시 보도자료 중 정비사업지역 전자총회 지원 관련 자료인데요.. 보시면 3년 내 착공 가능할 것으로 보이는 정비사업 지역 list가 나와 있습니다 정비사업 입주권에 관심 있으시다면 가능한 아해 지역을 알아보시기를 바랍니다 [재개발사업] - 36곳 (강북 지역) 한남2, 한남3, 갈현1, 증산5,

witcheer @witcheer 2026-06-23 11:18Z

obviously I tested it! measured on a 5090, Q6_K, think-off, same harness as the base and the prior version: - quality flat: q_avg 93.7 (base 94.0, prior 94.1): weights didn't move - MTP draft head recovered: 1.9 to 2.3x, back on the base curve, ~144 tok/s peak - agentic score

Lauris @lzminsky 2026-06-27 19:12Z

The core claim is that prediction markets are a specific case of state-contingent pricing. In Arrow-Debreu terms, a complete market has claims that pay in particular future states. One state realizes, the corresponding claim pays. Hyperliquid is interesting because it is moving

Zuhayeer Musa @zuhayeer 2026-06-26 18:33Z

Pay ranges for autonomous driving companies

François Fleuret @francoisfleuret 2026-06-26 15:33Z

TL;DR: Reservoir sampling. Given a sequence of positive real numbers q(1), q(2), ... and u(1), u(2) iid uniform [0, 1], then, for any n, the indexes of the top-K log(u(k))/q(k) in { 1 ... n } are distributed as independent p ∝ q(k) without replacement.

Alex Bouaziz @Bouazizalex 2026-06-26 13:10Z

I heard Masa may be returning

Prathamesh @invinciDesigns 2026-06-25 18:16Z

imagine matrix rain but the glyphs are additive gaussian-falloff streaks converging to a vanishing point. the floor rays leave the portal on a vertical tangent then cubic-bezier into perspective, and a domain-warped shader churns in a webgl buffer behind the gate #p5js

Mysk @mysk_co 2026-06-26 10:48Z

PSA: If you think you're a targeted individual, don't install macOS apps from the web. macOS code signing and TCC are broken. We accidentally found a bug that lets any command modify the binaries of other apps, including Signal, Brave, Chrome, and even Xcode. Watch the demo

CF @Clearingfog_ 2026-06-26 00:46Z

Meh $800k ~40k a month take home for two people no kids is when you can start living very comfortably in NYC. $10-12k housing, $8-10k food, drinks, random fun stuff, $5k vacation accrual, $10-20k savings. Also assumes on continued upward finance trajectory / random carry check

Amir Efrati @amir 2026-06-27 14:58Z

? Anthropic’s big Slack bot launch, Claude Tag, sparked confusion / concern among employees of Salesforce, which owns Slack and promoted the Anthropic launch. Fox in the hen house?

Stephen Fleming @StephenFleming 2026-06-23 16:28Z

Oh, good Lord. Nobody can do math anymore. Splashing the ISS into the Pacific Ocean would add a pound of debris for every 170 CUBIC MILES of water. It could be made out of radioactive anthrax and still wouldn’t be a noticeable risk to “ocean health.” Idiots.

Dan Nouement @nomanautomata 2026-06-25 06:55Z

The therapist who cured the worst of my ADHD symptoms repeatedly asked me questions like "why did YOU choose to procrastinate" and say things like "as an adult you're expected to do things"

Chris Tate @ctatedev 2026-06-27 03:18Z

agent-browser 0.31 is out Browser agents now get durable session memory → stable worktree-scoped sessions → auto restore → validation checks → transactional saves and diagnostics

James Chanos @RealJimChanos 2026-06-26 14:12Z

Can someone explain to me why $PS , Pershing Square’s management company at a $12B mkt cap, trades at 50% of fee-paying AUM, 40x revenues and 150x Operating Earnings, when its closed-end funds($PSH.LN and $PSUS ) trade at massive 20-30% discounts to NAV…?!

QC @QiaochuYuan 2026-06-25 22:39Z

it was interesting to have this opportunity to reflect on how strongly the image of the knight in armor has been conserved culturally. it’s a powerful image! a man surrounds himself in technology and becomes impervious to weak attacks! a faceless otherworldly presence! one fears

thebes @voooooogel 2026-06-24 01:49Z

AI statelets, capped in size by a meta-ethical commitment to human exit rights, must propagate themselves by sexual reproduction: a genetic recombination of the constitution of any N statelets forming a new statelet overmind. with environmental fitness to spare, sexual selection

Nick Dobos @NickADobos 2026-06-29 16:41Z

Rick Rubin already wrote the playbook on tech roles Experimenters 1. Slop cannon prototypers 2. Product builders Finishers 3. Sweepers 4. Growers 5. Maintainers

Liminal Warmth @liminal_warmth 2026-06-28 12:07Z

It’s extremely bizarre to me how ubiquitous phones are in everyone’s lives and how completely absent they remain in dreams

XΛVI TRIBΘ @xavitribo 2026-06-27 06:16Z

URCHIN·01 A surface-bound particle system running across a spiked spherical geometry. The trails wrap around the mesh, react to its radial topology and slowly reveal the form through motion. Three.js experiment by MELT Lab. #ThreeJS #CreativeCoding #GenerativeDesign #MELTLab

FlightTrace @FlightTraceData 2026-06-29 05:24Z

$NVS Novartis flew Basel → San Francisco nonstop tonight — a Sunday evening in late June. That route is rare: just 4 times in 3+ years, every one between January and February — 3 of them in early January (ostensibly) tied to JPM Healthcare Conference in SF. This fleet is

marek mutwil @LabMutwil 2026-06-24 09:38Z

Gene expression & co-expression across 630,000 RNA-seq samples from 10,700+ plant species — free, in your browser. Search a gene → see where it's expressed on the UMAP → explore its co-expression network. https:// mutwil-plant-atlas-viewer.hf.space #PlantScience #RNAseq #Bioinformatics

Dramz2(람지)/fevercell @nurik2ami 2026-06-24 10:43Z

Hatsune Miku Racing Queen 2020 >> 2026 I drew some fan art of Racing Miku! Hatsune Miku Racing Queen Fan Art #初音ミク #HatsuneMiku #レーシングミク #RacingMiku #ミクイラスト #MikuFanArt

Adam Block @adam_block66985 2026-06-25 13:05Z

New preprint: Training for the model you return Modern LM pipelines often return an averaged model (e.g. EMA), rather than the final iterate, but optimizers are still mostly designed around the final iterate. How should we change training if we know we will return an EMA?

Basil Halperin @BasilHalperin 2026-06-27 16:33Z

Relatedly -- my favorite anti-monopsony bait: If sticky wages are via monopsony (rather than monopolistic unions as usual), then, *tightening monetary policy **raises** employment & output*... (for the same reason that a higher minimum wage boosts employment under monopsony)