Search the archive

Most AI agents never reach production. Agent Bricks by Databricks

fixes that by helping you build agents that actually work — accurate, reliable and grounded in your data.

See why it's worth your time

SPACEX IN TALKS FOR $800 BILLION VALUATION AHEAD OF POTENTIAL 2026 IPO

SpaceX is holding a secondary share sale that may value it at $800 billion. The company is also weighing a potential initial public offering in 2026. SpaceX investors have been waiting for an IPO for years. The company has grown into an essential service for the US government. It also provides broadband internet around the world.

THE INSIDE STORY OF HOW NETFLIX WON THE WARNER BROS. AUCTION

Netflix outmaneuvered Paramount to buy Warner Bros. in a deal that may change Hollywood forever. The deal still needs to pass regulatory scrutiny before it can pass, and there is plenty of opposition against the deal. Paramount could still launch a hostile bid. This article tells the story of the acquisition based on conversations with more than a dozen individuals directly involved in the process.

AFTER NEURALINK, MAX HODAK IS BUILDING SOMETHING EVEN WILDER

Max Hodak co-founded Neuralink with Elon Musk in 2016. He served as its president and essentially ran day-to-day operations. In 2021, he left and co-founded Science Corp., a brain-computer interface company that aims to build technologies that can reshape human consciousness. Its first product is a procedure that restores 'form vision' to people with advanced macular degeneration through a rice grain-sized computer chip implanted directly in the retina.

THE BIG BLIND

Zanskar is a company that uses advanced AI models to prospect for geothermal sites. Its AI models recently discovered a massive underground heat source in western Nevada. This is the third such finding in Zanskar's short life. AI prospecting appears to be a game-changer for geothermal energy, a market that many had previously written off because finding heat sources was so challenging.

Schedule a demo to learn more →

HOW I KEEP UP WITH AI-GENERATED PRS

AI can generate code really fast, but that code still needs to be reviewed. Automated PR tools can help, but they're not sufficient. This article presents a workflow that can cut the time it takes to review AI-generated PRs significantly. The workflow uses AI to summarize changes and identify potential issues while allowing the developer to stay in control of the final output.

ESTIMATES – A NECESSARY EVIL?

Developers find it difficult to estimate. A lot of the time, it can be difficult to see why the estimate is even important. Product owners see estimates as a crucial piece of the puzzle that helps in planning the short and long-term life cycle of a product. It is best to work in a way that allows developers to be transparent and continuously communicate updated estimates as development progresses.

ANOTHER DEEPSEEK MOMENT

ZTE's Nubia M153 is a smartphone that runs ByteDance's DoubaoAI agent at the OS level. The agent has complete control over the phone. It can see the UI, choose/download apps, tap/type, make calls, and run multi-step task chains. The Doubao model is a massive, sparse Mixture of Experts model with full text and vision support. This thread contains several videos that show what the phone is capable of.

I WASTED 8 YEARS OF MY LIFE IN CRYPTO

Bitcoin will probably hit a million one day, but that is irrelevant to the financial games the industry is churning out. The industry mentality is incredibly toxic and will likely lead to the long-term collapse of social mobility for the younger generation. It is up to us to find the courage to resist the pull of worthless games. Sometimes, it is better to be right than to make money.

productivity. Get your copy from Miro

APPLE ROCKED BY EXECUTIVE DEPARTURES, WITH CHIP CHIEF AT RISK OF LEAVING NEXT

Johny Srouji, senior vice president of hardware technologies and one of Apple's most respected executives, recently told Tim Cook that he is seriously considering leaving in the near future.

X CUTS OFF THE EUROPEAN COMMISSION'S AD ACCOUNT AFTER BEING FINED €120 MILLION

X was fined about $140 million for violating the Digital Services Act.

WORKING FROM HOME IS HARMING YOUNG EMPLOYEES. THEY'RE STARTING TO SEE THAT

Some younger workers have decided to spend more time in the office due to feelings of isolation, a desire for more mentoring and feedback, and to improve the odds of a promotion.

OPENAI'S RESPONSE TO GOOGLE GEMINI: CHATGPT 5.2 LAUNCH TIPPED FOR NEXT WEEK

GPT-5.2 could help OpenAI close the gap created after Gemini 3's strong debut.

THE BINDING PROBLEM

Answering the binding problem will require advanced neural interfaces that allow scientists to experimentally access the brain to run experiments.

TIKTOK ROLLS OUT NEARBY FEED TO BRING YOU MORE LOCAL CONTENT

TikTok's Nearby Feed is only being rolled out to a select few countries for now, including France, Germany, Italy, and the UK.

3️⃣ Tooling evaluation framework + scorecard

Tackle problems early in the lifecycle by accelerating your

prototyping strategy. Get the guide to learn how.

META ACQUIRES AI DEVICE STARTUP LIMITLESS

Meta has acquired Limitless, a startup known for its AI pendant that records conversations. Limitless will cease hardware sales but continue to offer support for a year. Its team will focus on supporting Meta's existing AR and AI wearables rather than launching new products. The acquisition comes amid rising competition in AI hardware as major players like Meta expand their device offerings.

GOOGLE SET TO LAUNCH GEMINI NANO BANANA 2 FLASH

Google plans to launch the Gemini Nano Banana 2 Flash model, which matches the performance of the Nano Banana 2 Pro, at a lower cost. The move targets a broader user base by providing high-quality outputs suitable for large-scale deployments. Code references suggest an announcement is imminent, aligning with Google's strategy of model tiering for cost optimization.

NEXT CHATGPT UPGRADE IMMINENT FOLLOWING ‘CODE RED' DECLARATION

OpenAI could be releasing GPT-5.2 as early as tomorrow. The update will focus on speed, reliability, and customizability. It is aimed at closing the gap that Google created with the release of Gemini 3 last month. OpenAI had declared a 'code red' in response to Gemini 3's release.

ARC PRIZE 2025 RESULTS & ANALYSIS

The ARC Prize 2025 Score and Paper winners have been announced. 1,455 teams submitted 15,154 entries for the ARC Prize 2025. All ARC Prize 2025 winning solutions and papers were open-source. This post takes a look at the winners and how ARC-AGI will push AI reasoning further in 2026.

TRAVERSING THE FRONTIER OF SUPERINTELLIGENCE

Poetiq has established a new state-of-the-art on the ARC-AGI-1 and 2 benchmarks. Its approach of building intelligence on top of any model allows it to integrate new models within hours of their release. The company's core meta-systems produce optimized agents to automate the extraction of knowledge for hard tasks that require complex reasoning. This article looks at how Poetiq's team built its systems.

ANNOUNCING RNJ-1: BUILDING INSTRUMENTS OF INTELLIGENCE

Rnj-1 is a world-class pair of base and instruction-tuned large language models. The 8B model roughly follows the open-source Gemma 3 architecture. The Rnj-1 Base and Instruct models compare favorably against similarly sized open weight models. This post summarizes the key capabilities of the models and covers the background behind their development.

TITANS + MIRAS: HELPING AI HAVE LONG-TERM MEMORY

Google Research has released two new papers introducing an architecture (Titans) and a theoretical blueprint (MIRAS) that combine the speed of recurrent neural networks with the accuracy of transformers. Together, Titans and MIRAS advance the concept of test-time memorization, the ability of an AI model to maintain long-term memory. They introduce a meaningful shift toward real-time adaptation. The architecture actively learns and updates its own parameters as data streams in, enabling the model to incorporate new, specific details into its core knowledge instantly.

CUTILE PYTHON

cuTile Python is a programming language for writing parallel kernels for NVIDIA GPUs. It automatically leverages advanced hardware capabilities while providing portability across different NVIDIA GPU architectures. cuTile enables the latest hardware features without requiring code changes.

COMPUTER-SCIENCE REINFORCEMENT LEARNING GOT REWARDS WRONG

In reinforcement learning, the reward is part of the agent, not part of the environment. Once we start thinking in terms of the reward translation as being part of the agent, we can also start thinking of ways in which the agent can affect the reward translation mechanism. It could be dynamic and change across time, be tied to policy, or even learned. The shift in perspective creates many possible options, just by changing the notation - because notation matters.

GEMINI 3 PRO ADVANCES VISUAL REASONING

Google's Gemini 3 Pro set new benchmarks in multimodal vision tasks, excelling at document, screen, spatial, and video understanding with enhanced visual reasoning capabilities.

performance—enterprise, neocloud, and beyond Learn how it works

GOOGLE PLANS TO INTEGRATE OPAL AGENT BUILDER INTO GEMINI

Google is exploring integrating Opal's agent-based automation into Gemini, enhancing direct access to advanced AI tools.

TENSOR PARALLELISM IN TRANSFORMERS

A concise breakdown of how tensor parallelism splits Multi-Head Attention and Feed-Forward layers across GPUs to enable large-scale training of transformer models.

THERE'S GOT TO BE A BETTER WAY!

Reformist reinforcement learning seems to have found a strong niche in fine-tuning language models, but the techniques you get from reinforcement learning are almost always bad.

CANADA'S CRYPTO TAX CRACKDOWN

Canada's tax agency (CRA) has pulled in over $100M CAD (~$72M) from crypto audits in the last three years, but hasn't brought a criminal case to court since 2020. Although it estimates ~40% of users on platforms that it sees are non-compliant or high-risk, the CRA admits it still can't reliably identify most crypto users. As part of only its second-ever crypto data dragnet, the CRA forced Dapper Labs to hand over info on 2,500 users, following an earlier order against Coinsquare. The gap between money collected and cases prosecuted stems from cross-border complexity, scarce evidence, and slow international cooperation, even as Canada gears up a new financial-crimes agency and continues big AML fines via FINTRAC.

POLAND LEFT AS MICA OUTLIER IN THE EU

Poland's parliament failed to override the president's veto of its MiCA-alignment bill, leaving the country as the only EU member without a domestic MiCA framework. Germany, Malta, the Netherlands, and Lithuania are already issuing MiCA licenses. However, Poland's crypto market is booming anyway. It ranks 8th in Europe by on-chain value, with >50% YoY volume growth and ~7.9M people using crypto. Meanwhile, Italy is warning firms they must be fully MiCA-compliant or shut down by December 30, and Brussels is exploring a future model where a single EU-level supervisor regulates exchanges across the bloc, which could eventually reduce the impact of national holdouts like Poland.

METAMASK INTEGRATES POLYMARKET

MetaMask has integrated Polymarket into its mobile app, enabling one-tap funding with any token on any EVM chain and MetaMask Rewards points toward a potential MASK airdrop as Consensys reportedly plans an IPO. The integration provides a new access point for Polymarket, which recently secured CFTC approval to re-enter the US and is courting a $15 billion valuation after receiving $2 billion in strategic funding from NYSE parent ICE at a $9 billion valuation.

KRAKEN LAUNCHES HIGH-TOUCH VIP PROGRAM

Kraken launched an invitation-only VIP program that requires $10 million average balances or $80 million annual trading volume, pairing ultra-high-net-worth clients with dedicated relationship managers that provide 24/7 support, subject-matter expert access, and exclusive Formula 1, football, and cultural experiences ahead of its early 2026 US IPO. The high-touch service competes with Binance's recent "prestige" offering for clients up to $10 million net worth as exchanges fill the wealth management gap for family offices and sophisticated traders.

BITCOIN FACES JAPAN RATE HIKE: DEBUNKING THE YEN CARRY TRADE

Fears of a yen carry trade unwind from the Bank of Japan's expected 0.75% rate hike are overblown as the move is already priced into Japanese government bond yields near multi-decade highs, and speculators hold net long yen positions, limiting panic buying potential, contrasting with August 2024's bearish positioning that triggered market chaos. The yield differential between Japan's 0.75% and the US' 3.75% remains wide enough to favor American assets and discourage mass carry trade unwinding, with BOJ staying the most dovish major central bank. The real risk lies in Japanese tightening sustaining elevated US Treasury yields that counter Fed rate cuts, dampening global risk appetite by raising borrowing costs, rather than a sudden yen surge, while Trump's fiscal expansion push could stoke debt fears, lifting bond yields further.

MARKET ECHOES EARLY 2022 AS ONCHAIN STRESS MOUNTS

NYT'S “BANK DEPOSIT VS STABLECOIN WALLET” GRAPHIC BACKLASH

A New York Times infographic contrasts bank deposits (insured, interest-bearing, with legal rights) against stablecoin balances (uninsured, issuer keeps yield, weaker legal claims). However, this is a stacked comparison as banks historically share very little of the yield they earn, many countries have weak or no deposit insurance, and large depositors are just unsecured creditors of leveraged banks. In crypto, users can move stablecoins into DeFi or tokenized T-bill products to capture much higher yields directly, and self-custody avoids many of the freezes and capital controls that can hit bank accounts.

CRYPTO IS GOING MAINSTREAM

Crypto is quietly integrating as a secure communication layer between applications rather than through dominant L1s like Ethereum or Solana. Stablecoins already process $9 trillion adjusted annual volume, rivaling Visa and PayPal by representing fiat balances on interoperable blockchain transport. Apps like PayPal and LINE Pay are operating as independent L2s posting ordered transactions to L1s using erasure coding, where each validator handles constant data loads regardless of scale. Monad excitement reflects outdated mental models, as fast finality became table stakes and developer lock-in to monolithic VMs contradicts a decade of experience showing easy EVM portability, while unified liquidity emerges from universal base layers enabling secure communication across chains rather than single blockchains attempting everything.

ETHEREUM ‘KILLER' VS ETHEREUM L2 REALITY

From 2019 to 2024, Solana marketed itself as the scalable, monolithic alternative to “band-aid” L2s, but by 2024 to 2025, Base was leading L2s in TVL growth, daily active users, and meme/creator activity, even overtaking Solana in new token launches and associated DEX volume in August.

ETHEREUM FEES & STABLECOIN LIQUIDITY

Median transaction fees on Ethereum are at all-time lows, even as dollar stablecoin supply on Ethereum hits all-time highs.

ZKSYNC LITE DEPRECATION IN 2026

ZKsync is planning to shut down ZKsync Lite (its original 2020-era ZK rollup) sometime in 2026, calling it a completed “proof-of-concept” that paved the way for newer products like ZKsync Era, ZK Stack chains, and Prividiums.