Search the archive

from Wispr Flow. No meeting bots required. Use it for free

GOOGLE ACCELERATOR AGENTS FOR TPU DEVELOPMENT

Google's Accelerator Agents use Gemini to help developers migrate PyTorch workloads to JAX and optimize custom kernels for Google Cloud TPUs. The toolkit includes MaxCode for model conversion and MaxKernel for writing, porting, profiling, and debugging Pallas kernels.

QWEN-DRIVE

Qwen-Drive is a project that aims to create a Vision-Language Foundation model for autonomous driving. The project uses a staged training strategy that integrates perception, language, and planning objectives. The model achieves specialized driving competence while retaining broad visual understanding and instruction-following capabilities. A GPU with 24 GB+ of memory is recommended.

HIP-AGENT: A HARNESS THAT FITS IN THE PROMPT

hip-agent is a small agent harness designed for agents. The configuration is environment variables, actions are shell commands, and a subagent is a child process. The rest is handled by existing protocols and formats. The core loop is about 200 lines of Python and a module for the Codex API.

PRODUCT MANAGER, APPLIED AI AT TLDR ($200K BASE + $60K BONUS, FULLY REMOTE)

TLDR is hiring its first PM to help build the agent-first operating layer used across the company. We're looking for a builder who has shipped real products/systems with LLMs. Click here to learn more [19].

THE EDUCATION OF A DOOMER

Automation has been good, but AGI may make humans economically useless, which will have many consequences. It's clear that AI capabilities are growing far faster than our ability to control or understand them. In the future, people will likely hand over control to AI voluntarily, as it will be the logical choice. However, before that happens, there are several signs that people are already giving up thinking.

MACHINES THAT THINK: EMBODIED INTELLIGENCE

Vision and language models are making strides, but embodied AI in robotics struggles due to sparse, costly training data for manipulation tasks. Robotics excels where tasks are fixed, yet general-purpose robots falter, evident in low performance on benchmarks like Libero. Startups should focus on narrow, instrumented deployments. Innovation lies in converting existing robot data into training insights. Ensuring reliability in physical AI systems is paramount.

THE CHASM: THE SHAPE OF UNFINISHED AI CODEBASES

AI codebases exhibit a unique, unpredictable pattern where programs appear polished but often hide deep, hidden issues leading to significant failures outside controlled demos. Unlike human-authored programs where gaps surface predictably, AI-generated software misleads with false performance tests or incomplete features, resulting in the need for frequent rewrites to address these deep-rooted flaws. Understanding and anticipating the underlying shape of these unfinished AI codebases is crucial to navigating and mitigating the challenges they present.

COSINE SIMILARITY IS NOT A SAFETY PROPERTY

Cosine has no notion of truth, authority, or provenance.

ARM'S C2-ULTRA, G2-ULTRA NX, AND CSS N4 IP

Arm's upcoming C2-Ultra CPU and G2-Ultra NX GPU IP will see broad use across flagship phones.

BYTEDANCE IS PREPARING A REAL-TIME SPATIAL VIDEO MODEL UNDER ZHANG YIMING

ByteDance, led by Zhang Yiming, plans to launch an AI model for real-time spatial video generation, potentially as early as next month.

GOOGLE TESTS AI-POWERED CONTRAIL AVOIDANCE ON LONG-HAUL FLIGHTS

Google is trialing contrail avoidance for ultra-long-haul flights in the Asia-Pacific region.

LOVABLE LAUNCHES DRAFTS FOR PARALLEL APP EXPERIMENTATION

Drafts allow tech teams to experiment with project changes without affecting live apps, enabling parallel version exploration.

DBS/CITI COMPLETE FIRST WEEKEND CROSS-BORDER USD PAYMENT

DBS and Citi completed a Singapore-to-US USD payment over a weekend using tokenized deposits via Swift's Digital Ledger, taking minutes versus up to two business days for traditional cross-border transfers. The move removes weekend/time-zone delays for institutions operating across jurisdictions. Citi joined Swift's 24/7 tokenized payments pilot in July and is also part of a group of major US banks building a separate tokenized deposit network through The Clearing House, targeting a 2027 launch.

HANWHA BUILDS TOKENIZED SECURITIES PLATFORM ON AVALANCHE

South Korean brokerage Hanwha Investment & Securities reportedly built a tokenized securities platform (started in 2025 with FairSquare Lab) running on Avalanche and Hyperledger Besu, ahead of South Korea's securities tokenization law taking effect in February. Regulators plan a phased rollout starting with private money market funds, bonds, and unlisted stocks, eventually expanding to public securities and stablecoin settlement rails. Hanwha is also Securitize's largest shareholder (9.6% stake) and separately invested $22.3M in Canton Network operator Digital Asset.

ROWDAS (EIP-8371): DISTRIBUTED BLOB RECONSTRUCTION, MEASURED

EIP-8371 (RowDAS) proposes adding 128 row-based subnets to Ethereum's PeerDAS architecture to eliminate the redundant blob reconstruction work that every qualifying node performs on its own. Under the present design, an estimated 2,000+ nodes each spend 5.18 seconds of CPU per slot reconstructing 32 blobs. RowDAS distributes that load, cutting per-reconstructor time to 40 milliseconds and network-wide CPU from 10,400 to 81 seconds at 32-blob scale. The design uses a two-dimensional cell matrix where column subnets retain custody authority while row subnets pool reconstruction, allowing ordinary nodes to reach the erasure-coding threshold without high-stake infrastructure. A Prysm prototype with real KZG cryptography exists, though mainnet-scale testing is pending, and a duty-sharded fallback variant preserves 91% of CPU savings for staged rollout before the full row layer activates.

WYOMING PUTS ITS STABLECOIN RESERVES ONCHAIN WITH UNDER $1 MILLION OUTSTANDING

Wyoming's Stable Token Commission adopted Chainlink Proof of Reserve on September 2 as the exclusive verification layer for the Frontier Stable Token (FRNT), routing reserve and supply balances audited by The Network Firm to an onchain feed. FRNT's outstanding supply of 967,948 tokens priced at $0.9988 each represents 0.2% of the 508.7 million token base case the Commission projected to state lawmakers in December, with just $160 in 24-hour trading volume occurring on Kraken. The Proof of Reserve Secure Mint component, which would programmatically block minting without verified reserve backing, remains unimplemented. The integration is Wyoming's second Chainlink partnership in a month, following an August 18 migration from LayerZero to Chainlink CCIP that the Commission attributed to security review concerns and LayerZero's disclosure practices.

WHO KILLED THE CRYPTOANARCHIST?

Cryptoanarchy, the movement centered on data privacy, permissionless systems, and trust minimization, retreated during the 2020-2022 bull cycle as crypto companies prioritized institutional capital and mainstream UX over ideological principles. Each value the movement abandoned has since been validated by AI-era behavior: users now hide camera rolls from Instagram and stockpile Mac Minis to host open-source models, while businesses treat every networked system as an attack surface for misaligned agents. Most original cryptoanarchists have either exited the space or reframed their work for institutional audiences, leaving the movement hollowed out precisely when broader society began internalizing its core concerns. A revival is possible under new banners, with AI-driven threat models replacing the 2008 financial crisis as the animating fear.

THE SHAPE OF ETHEREUM: A SIX-YEAR STUDY OF TOPOLOGICAL ANOMALIES

Researchers applied Topological Data Analysis to six years of Ethereum transaction data (2020-2025), segmenting daily activity into four layers and using Wasserstein distance to quantify structural changes between consecutive days. The pipeline flagged 86 anomalous periods, with 85% corresponding to documented events including Black Thursday liquidation cascades, DeFi repositioning two days after the Ukraine invasion, and post-Bybit hack fund dispersal across bridges. Unsupervised change-point detection identified six structural breaks tied to identifiable catalysts, and preliminary results show TDA features improve 7-day ETH volatility forecasts beyond GARCH baselines. Layer composition shifted from governance-dominated anomalies in 2024 (11 of 16 events) toward plain ETH transfers in 2025, though the absence of a labeled ground-truth dataset limits validation of which events structurally alter chain topology.

COLDCARD EXPLOITER MOVES 45% OF WAVE 3 STOLEN FUNDS

The attacker behind the Coldcard hardware wallet exploit (a 2021 firmware bug that weakened seed randomness) has moved 45% of Wave 3 stolen bitcoin, per Galaxy Research, swapping some to ETH via THORChain, then running funds through CoinJoin. About $7.8M has been spent so far, moving stolen vaults largest-first. 82% of all exploited funds remain in attacker-controlled addresses. Total identified losses across all waves now stand around 1,806 BTC (~$143.9M), with Galaxy flagging a possible Wave 4.

MAPPING ONCHAIN BTC CREDIT SOLUTIONS

Surveys four approaches to unlocking bitcoin-backed credit without a fully taxable wrap: Anchorage's Atlas (a legal collateral-agent structure mainly for CeFi lenders, thin smart-contract layer), Komainu (a Nomura/Ledger/CoinShares JV using a legal "Notice of Exclusive Control" and Liquid Network settlement), Babylon Labs (fully non-custodial, BTC never leaves its own chain via Taproot dual-spending paths), and Concrete's AssetCX (BitGo custody plus a real 1:1 ERC-20 synthetic usable as DeFi collateral). Concrete is the most "DeFi-native" since it's the only one built around an actual onchain token rather than a legal wrapper.

MENA CRYPTO VOLUME HITS $350B

Middle East/North Africa on-chain transaction volume grew from ~$100B in 2022 to ~$350B in 2025-2026.

BANKS BECOMING DEFI'S REAL COMPETITORS

Recent US bank charter approvals signal banks are about to gain the same programmability and 24/7 settlement DeFi protocols have relied on as differentiators.

CRYPTO CARD UPDATE ROUNDUP

Krak now requires a €200 average monthly balance for a free physical card.

SATOSHI-ERA BITCOIN WAKES AFTER 16 YEARS OF DORMANCY AS 600 BTC MOVES

600 BTC (~$48 million) transferred from 12 addresses dormant for over 16 years, each holding proceeds from a single block reward mined in March 2010 at the original 50 BTC subsidy.

GOOGLE'S GEMINI SPARK CAN NOW MANAGE YOUR GOOGLE PHOTOS LIBRARY

Google's Gemini Spark can now edit images, curate albums, auto-create shared albums, and turn concert flyer photos into calendar entries inside Google Photos. Google says the rollout arrives over the coming weeks to eligible Gemini AI Pro and Ultra subscribers in the US, though international timing remains unannounced.

ATLAS: A WORLD MODEL FOR SPATIAL INTELLIGENCE

World Labs has introduced Atlas, a next-generation world model for spatial intelligence, pretrained from scratch as a multimodal autoregressive diffusion transformer spanning text, images, video, and 3D. Atlas generates up to one minute of camera-controlled 1440p video, reconstructs real-world scenes from just two or three images, and simulates space and time for robotics. It outperforms specialized 3D reconstruction models on benchmarks and will power future versions of Marble and other World Labs products, now entering early access with select partners.

GOOGLE CLOCK ROLLS OUT REDESIGN OF TIMERS WITH PRESETS ON ANDROID

Google Clock's redesigned Timers tab introduces one-tap presets for 1, 5, 10, and 15 minutes, along with a larger Start button and a cleaner input layout for creating custom timers. Multiple active timers now appear in a space-efficient grid instead of a scrolling list, while the countdown controls have also been reorganized. The update is rolling out via a server-side update for Google Clock version 9.1.

HOW OUR AGENTS BUILD ON-BRAND PAGES WITH DESIGN.MD

Vercel's design.md is a public file any agent can load to build on-brand pages outside the codebases where its product-design skill lives. The system pairs that guidance with a public stylesheet defining fixed classes and tokens, plus an evaluation loop that turns reviewer feedback into rules and deterministic checks. After more than 200 test runs, pages generated with design.md loaded showed 57% fewer known layout failures than those generated without it.

READING AND WRITING ARE INTERFACES, AND AI CAN REDUCE THEIR FRICTION

Criticizing all AI-assisted writing overlooks the fact that poor-quality, derivative content existed long before generative AI, and using AI does not automatically diminish the intellectual value of someone's work. The key distinction is whether AI removes incidental effort, such as editing or formatting, or replaces the critical thinking, analysis, and judgment that a task is meant to develop. Rather than focusing on whether AI was used, the more meaningful question is which parts of the thinking were delegated and which remained the responsibility of the person.

GOOD ISN'T GOOD ENOUGH: BEWARE THE DESIGN MIDDLE GROUND

As creative tools become more accessible, the greatest risk is not poor-quality work but settling for polished, predictable ideas that blend into an increasingly crowded middle ground. Distinctive design comes from resisting safe, incremental decisions, embracing experimentation, and making room for curiosity, intuition, and personal exploration beyond client constraints. The most memorable creative work is driven by originality and conviction rather than technical execution alone.

A DAILY DESIGN QUOTE

A daily design quote: every day, a thought-provoking quote from a designer the site's creator admires.

GENERATIVE AI TOOLS AND DIGITAL ASSETS FOR CREATORS

Artlist blends premium assets with generative AI for video and image creation. Unleash your creativity with music, SFX, footage, voiceover, and AI tools.

THE MAC-NATIVE TOOLKIT FOR SCREENSHOTS, RECORDINGS, AND COLLABORATION

CleanShot X is a screenshot and screen recording app for Mac with a built-in annotation tool, Cloud uploading, scrolling capture, and a lot more.

HOW TO SCALE DESIGN TOKENS

Design token systems that work fine at 50 tokens often break at 500, typically because no one defined who owns changes, renames, or conflicting edits. Separating raw primitive values from semantic aliases lets a theme swap become a one-line change instead of a component-by-component hex-code audit, enforced with a simple linter rule. Treating renames as versioned, deprecation-windowed changes and generating every platform output from one source in CI, with a WCAG contrast check before shipping, keeps the system scalable.

SKOGEN BORROWS FROM SCOUTING, NOT TECH, FOR GRIDSCOUT'S IDENTITY

Skogen created a new identity for GridScout that positions the utility technology company as “a reliable colleague on every pole,” balancing technical credibility with a warm, approachable character for everyone from field crews to executives. The redesign replaces a literal woodpecker logo with a more abstract symbol, introduces a scout-inspired visual system with badges, rugged typography, and workwear-inspired colors, and reinforces the brand through authentic photography and a Norwegian-language launch film. The cohesive identity is designed to support GridScout's growth while remaining practical, recognizable, and trusted across every touchpoint.

PUBLIC ADDRESS REBRANDS CODE.ORG AS CODEAI TO TACKLE AI LITERACY

CodeAI has introduced a new brand identity to reflect its expanded focus beyond coding to AI, data science, and computational thinking, responding to the growing need for AI education in schools. Designed by Public Address, the identity deliberately avoids common AI clichés, instead presenting AI as code while combining technical elements with human-centered imagery to encourage critical thinking. The flexible, accessible system adapts to audiences ranging from students to policymakers while maintaining a consistent, future-focused visual identity.

AI RUINED POSTERS. THIS HUMAN DESIGNER IS FIXING THEM

AI-generated posters tend to look busy, information-dense, and especially unappetizing for food imagery, prompting designer Elizabeth Martin of Ello Design Studio to redesign real AI-made posters.

TURNING COMPUTATIONAL DESIGN INTO LIVING SCULPTURAL SKINS

3D Beast, a New York-based architecture and computational-design practice, translates architectural systems into wearable bio-engineered structures that read as part exoskeleton, part sculpture, part prototype.

OPENAI'S CONTROVERSIAL CHATGPT-6 ASTRA AD PAINTS A SAD VISION OF THE FUTURE

OpenAI has released GPT-6 Astra, an autonomous AI model capable of handling complex creative and technical workflows, prompting both excitement over its capabilities and concerns about safety, misuse, and its impact on creative professionals.

🗓️ Save your spot — Sept 16, 9AM PT / 12PM ET

HOW WELL DO AGENTS USE TEST AND VERIFICATION TECHNIQUES?

Across 26 testing conditions for a Rust Zstd task, no technique dramatically outperformed the others, and the default prompt performed above average. Fuzzing and property-based testing helped slightly in some high-effort runs, while TDD and several testing skills underperformed. Agents often followed the requested framework superficially, writing ordinary tests inside new tools or proving properties that did not improve correctness.

EXPLORING SPECULATIVE DECODING IN VLLM ON AMD GPUS

Speculative decoding lets vLLM verify multiple draft tokens in one target-model pass while preserving the target model's output behavior. Tests on AMD MI300X and MI355X GPUs show that throughput depends heavily on the drafting method, proposal length, model family, workload, and token acceptance rate, with setup and tuning guidance for MTP, EAGLE-3, DFlash, and DSpark.

FROM S3 TO GPU IN ONE COPY

Vortex streams compressed columnar data from S3 through host memory to the GPU with a single copy, avoiding disk staging and CPU decompression bottlenecks. Its layouts prune data before transfer, and its encodings support GPU-side decompression and computation, producing scans around 30 times faster than Parquet in the demo.

SIMPLE IS NOT SMALL

Small programs can still be tightly coupled, while large systems can present a simple interface. Simplicity comes from separating concerns and choosing representations that do not braid unrelated behavior together, not just minimizing lines of code.

PROGRAMMING IS ART

Some developers value programming as a creative practice, not just a means to ship products or earn money. Even if AI can generate code faster, people who love the process will keep writing code by hand because the act itself is the reward.

MAKING SOFTWARE HURTS NOW

AI adoption is eroding trust in open source, where maintainers now face suspicion over how contributions were produced while many workers cannot afford to opt out of mandated tools. The essay calls for more solidarity with people sharing public goods instead of directing justified anger at fellow developers.