OpenAI just revealed the first results from Jalapeño, its first custom AI inference chip, and the gains are huge. Across big models including DeepSeek R1 670B and Kimi K2.5 1T, it delivers 1.5-1.9x more AI work per watt, 1.7–3.6x lower end to end latency, and up to 4.1x higher performance for highly interactive workloads versus comparison systems. AI helped design the chip, enabling OpenAI to go from initial design to tapeout in only 9 months. AI generated implementations for selected workloads were already 1.5–1.8x faster than human expert written versions. AI designs better chips → better chips run AI faster and cheaper → stronger AI designs even better chips → repeat

JIO PLATFORMS' PROFIT RISES 13.7% IN ITS FIRST RESULTS SINCE ITS BLOCKBUSTER MUMBAI DEBUT Ambani's Jio posted consolidated net profit of 83.89 billion rupees for the September quarter, days after a record $3.2 billion IPO, per Reuters.

Tomorrow at #ESMO26#: MSK gastrointestinal medical oncologist @AndreaCercek will present the first results from the global AZUR1 trial of dostarlimab monotherapy in patients with dMMR/MSI-H locally advanced rectal cancer. 🗓️ Sunday, October 25 | 11:00 a.m. CET

Tokenized stocks went from pilot to product this month Bullish ran the first regulated tokenized equity trade on Solana. @RobinhoodCrypto added 100 more stock tokens to its chain. Crypto(.)com opened US equities and ETFs to European residents. @Securitize posted a record $4.3B in average tokenized assets in its first results since listing. Tokenized GameStop shares now trade on PancakeSwap as $GMEB.

🚨🇮🇱 Past the halfway mark, Likud's lead keeps shrinking: 20.8% against Yashar's 17.1%, with Bennett back in third With 55.4% of the vote counted, Netanyahu's party has lost almost three points since the first results, Eisenkot's Yashar has gained more than a point in the same stretch, and Bennett's Together alliance has nudged past the Joint List for third at 9.2%. The gap between the two front-runners was eight points at 4 a.m. and is under four now, with the big cities and the soldiers' envelopes still to come. Source: Central Elections Committee via Times of Israel, Al Jazeera / Writer: Daniel

Lido Poolside: Institutional Stop by on Thursday, September 17th, to discuss the latest institutional Ethereum staking developments and first results from the recent Lido Core upgrade. Featuring special guests from @Consensys and @Sharplink. More info:

What's happening with UK LLM is wild. People are giving up their time, sharing what they know, and actually building an LLM together. We've already got a council pilot, a shared test set and the first results coming in. We're pulling the testing together now. The local AI community in the UK is seriously strong. Mega seeing everyone get stuck in 💪🏻 If you are in the UK and want to participate, then please join in on Discord - Big thanks to everyone contributing ideas, tests and their time: 1O48576 AaronBedward AIAnd_I AIMystik AldakerA andydynamics atasrb AutomatedAgile BexheartsAI bitplane cepstrum9 Chronara_ai cryptosibbers dafzthomas danielkempe Dataphil3 DeanYates123 DeliriousJ51867 DerekColley_ Disclosure1990 dnixty dubsthedev Egly443 Fletch_K_ fran0245 fxdstudios GasPoweredCat halfpennymac HouMuza iamkevchilds ImperialisVox javierbits Jayeljay2 jbennieme jtdavies KixLabs_UK kristiannaaron Lewislee215213 Luc_Gibson Luke85953209 M_LeninistSco1 machinegenie magnvsltd Mark_Edgeller MartinSWDev mikenott mrfallah p0sts1ngular1ty P2PJim PapaRichens PassingPixels photo_png raftomas RicDerby sennnenac sherlock_comms ShockerLocker socloapp ssiliconbg subkleo TechnicalCooper TheLabJam ThomasKleitc wb1936

Jev really inspired me to build SUPER fast browser agents without sacrificing accuracy. I gave Codex access to vLLM on 2×B300 and let it change everything from the harness to inference. The constrained optimization: > min end-to-end task time > s.t. score ≥ baseline Caching, thinking, action batching, inference. Any part of the pipeline is fair game. First results below on 12 local form tasks. The goal: 10× faster on long tasks.

Catching up with the chain after being offline will see massive speedups with the upcoming Glamsterdam hard fork. Here's how it works: Today, nodes that come back online enter what's called "full-sync" mode. This means fetching blocks from peers and executing them. This guarantees that nodes verify each state transition and produce transaction receipts (=outputs that tell you "how" your tx executed, the amount of gas it used, logs, etc). With EIP-7928: Block-Level Access Lists (BALs), each block comes with a BAL (=state diff) attached, allowing nodes to apply the changes listed in each BAL to their local state and arrive at the correct post-state, validated against the state root in the block header. This has already been useful for snap sync, and Glamsterdam ships snap/2, improving upon classical snap sync by increasing robustness and speed. It turns out, it's also useful for catching up with the chain after being offline for a while. Instead of entering full-sync mode, nodes can download blocks and BALs from peers and then use the BALs to derive the latest finalized state, and the state diffs between blocks needed for reorgs. Instead of re-executing all transactions, they simply apply the BALs. This works fine for most nodes but some might also want to have the receipts for finalized transactions. In such cases, nodes can either re-execute or fetch receipts from peers. This is safe because execution is only skipped for finalized blocks, as reported by the attached CL client, and it's fast. Nethermind has been experimenting with this, and the first results are pretty wild: 2,000 blocks synced in 40 seconds instead of 76 minutes. That's 115x faster. These are early devnet benchmarks, so take them with a grain of salt, but this is very much the kind of improvement that could make life easier for RPCs, full-node/desktop wallets. Check out @Nethermind 's PR: