Subscribe
Briefs

Daily roundup

Models, tools, enterprise AI, research, policy, product. Distilled to what changes positioning.

Saturday, Jul 18, 2026

The Clank weekly synthesis, briefs published July 13 through July 18 2026 weekly-readapplied-ai

Five days of AI news, zero frontier model launches, and the hardware and applied layers ate the entire tape

The week of July 13 to 18 produced roughly a dozen material AI stories. Not one of them was a foundation-model announcement. Samsung finally cleared Nvidia's HBM4 qualification. AMD moved MI450 into GA on a price-per-token framing rather than a per-FLOP one. TSMC printed a June revenue record. Boston Dynamics disclosed 1,000 Stretch units. 1X shipped Neo into homes with a teleoperations footnote. Unitree cleared a $619 million Shanghai STAR listing. Abridge crossed $500 million ARR. Cursor background agents went GA at the tier that just priced Anysphere at $3 billion. Google Cloud Next spun up Gemini Enterprise. Claude Corps applications closed. It is the first full week since early 2023 where 'model capability' was not the headline, and the shape of what replaced it is the actual story.

Friday, Jul 17, 2026

Boston Dynamics, 1X Technologies, and Unitree Robotics disclosures, July 14 to 17 2026 humanoid-robotsboston-dynamics

Three humanoid disclosures in five days triangulate one argument, and none of them are about the humanoid

The week's robotics headlines lined up in a pattern the category has been avoiding since 2024: Unitree cleared for a $619 million Shanghai STAR IPO on July 14, 1X Neo's first Bay Area shipments on July 16 with an Oslo teleoperations disclosure baked into the launch blog, and Boston Dynamics confirming 1,000 Stretch units deployed on July 17. Three different companies, three different form factors, three different capital markets. The single argument all three land on is that humanoid revenue is not, at the moment, a humanoid business. It is a warehouse business, a policy-arbitrage business, and a subsidized-teleoperations business. Investors have been buying the humanoid story. Operators have been buying something else.

Thursday, Jul 16, 2026

Boston Dynamics press briefing / Reuters robotics coverage, July 16 2026 boston-dynamicsstretch

Boston Dynamics hit 1,000 Stretch deployments, and the boring warehouse robot is winning the humanoid race that isn't a humanoid race

Boston Dynamics disclosed Thursday that its Stretch warehouse robot, the trailer-unloading arm-on-a-mobile-base it started shipping in volume in 2023, crossed 1,000 deployed units at customer sites during the second quarter of 2026. The disclosure sits inside a routine hardware-refresh briefing about Stretch v2, which the company plans to start shipping in the fourth quarter with a bigger payload envelope and a longer battery duty cycle. The 1,000-unit number is roughly ten times any single humanoid maker's deployed base and roughly twenty times 1X, Figure, and Sanctuary combined.

Samsung Electronics investor release / The Korea Economic Daily / SemiAnalysis, July 16 2026 samsungsk-hynix

Samsung finally cleared Nvidia's HBM4 qualification, and the memory duopoly is a duopoly again

Samsung Electronics confirmed Thursday morning in Seoul that its 12-hi HBM4 stack has passed Nvidia's qualification process for use in the Blackwell Ultra 300 series, ending a two-year run in which Samsung was effectively locked out of the top Nvidia SKU. Initial shipments start in September, with volume ramp targeted for the fourth quarter. SK Hynix had absorbed nearly all of the HBM3E allocation and the majority of early HBM4 volume, which is a nice way of saying that the highest-margin memory in the industry has been a one-vendor market since 2024.

Abridge company blog / Rock Health quarterly funding report, July 16 2026 abridgeambient-scribing

Abridge crossed $500m ARR, and ambient scribing is quietly the largest applied-AI category in healthcare

Ambient clinical documentation vendor Abridge disclosed Thursday, inside a hiring blog post that was clearly written as a soft touch for its ongoing Series F process, that annualized recurring revenue crossed $500 million during the second quarter, up from a reported $350 million in April. The platform is installed at 187 US health systems covering more than 300,000 clinicians. The number matters less as a company data point and more as a category one. Ambient scribing is now the single largest applied-AI category in healthcare by revenue, larger than clinical decision support, imaging AI, and revenue-cycle AI combined.

Wednesday, Jul 15, 2026

1X Technologies blog / IEEE Spectrum, July 14 2026 1x-technologiesneo

1X shipped Neo to its first ten households, and the operators are calling in from Oslo

1X Technologies confirmed on Tuesday that the first ten Neo home humanoids have been delivered to preorder customers in the San Francisco Bay Area, kicking off a rolling rollout of the initial 500-unit batch through the third quarter. The company also published, in a blog post that took most of a day to work out how to read, that Neo units currently rely on remote human teleoperators for a majority of unstructured household tasks, with an Oslo-based operations team on shift to take control of individual robots on demand. Fully autonomous tasks at launch are limited to a fixed set: fetching a labeled item from a labeled location, loading and unloading a dishwasher of a supported model, and folding towels of a supported size. Everything else defaults to a human somewhere in Norway watching through the robot's eyes.

AMD investor relations / The Register / SemiAnalysis, July 15 2026 amdmi450

AMD's MI450 shipped, and the pitch is finally price per token, not price per FLOP

AMD announced general availability of the Instinct MI450 on Wednesday, with initial capacity committed to Microsoft Azure, Meta, Oracle Cloud Infrastructure, and a still-undisclosed sovereign-cloud customer in the Gulf. Peak throughput is 3.1 exaflops FP8 per eight-GPU node, roughly 1.5x MI350 and, per AMD's own comparison chart, within 15% of Nvidia GB300 NVL72 on a per-node basis. The interesting shift is how AMD priced it. The MI450 list-price envelope, per multiple order-book leaks, undercuts GB300 by 30 to 40% on a delivered-token basis for Llama-4 and Qwen-3 inference workloads. That is the first time AMD has led with the workload economics slide instead of the peak-FLOPS slide.

Anysphere blog / The Information, July 14 2026 cursoranysphere

Cursor launched a Background Agents tier at $200 a seat, and Anysphere quietly confirmed a $3B ARR run rate

Anysphere rolled out Cursor Background Agents on Tuesday as a new paid tier at $200 per seat per month, an order of magnitude above the existing $20 Pro plan. The feature spawns long-running autonomous agents on Anysphere-hosted compute that plan, edit, run tests, and open pull requests against a customer's repo without a human loop, capped at 40 concurrent agents per organization at launch. In the same post, CEO Michael Truell disclosed an annualized recurring revenue figure of $3 billion, up from the $500 million number reported in April 2025. That is the fastest ARR ramp in software history, by roughly a factor of two, and it is happening while every other coding-agent company (Cognition, Replit, Codeium, GitHub) is fighting for the same buyers with worse retention numbers.

Compiled from national AI strategy documents and public procurement disclosures, 2024-2026 sovereign-aigpu-supply

Sovereign AI stopped being a marketing slide and became a national budget line, and the arithmetic favors the same six countries

Nine governments have now committed direct capital to owning GPU capacity as national infrastructure. The public numbers add up to roughly forty billion dollars in announced spend through 2027, which sounds like a lot until you notice it is smaller than what Microsoft, Amazon, Google, and Meta have each spent on data center capex in a single quarter of 2026. The gap between the sovereign-AI pitch and the sovereign-AI arithmetic is where the next eighteen months of policy fights live.

Tuesday, Jul 14, 2026

TechCrunch / Apple v. OpenAI complaint filed N.D. Cal., July 10 2026 appleopenai

Apple is suing OpenAI over stolen trade secrets, and the giveaway is that OpenAI has a Chief Hardware Officer now

Apple filed suit in the Northern District of California on July 10 against OpenAI, its Chief Hardware Officer Tang Tan (previously Apple's VP of iPhone and Watch product design for 24 years), and senior systems electrical engineer Chang Liu (previously eight years at Apple). The complaint alleges systematic trade-secret theft: candidates asked to bring Apple components to interviews, departing employees coached on how to evade Apple security, a retained Apple laptop with confidential downloads, misrepresented authorization to use a proprietary metal-finishing technique. Apple wants an injunction, the materials back, and evidence preserved. The lawsuit is about a device, not a model, and the reason it matters is that Apple has now formally acknowledged that OpenAI is a hardware company.

TSMC investor relations / CNBC / Reuters / SEC Form 6-K, July 13 2026 tsmctaiwan-semiconductor

TSMC's June was the biggest revenue month in its history, and that number is the AI capex conversation ending

Taiwan Semiconductor posted T$442.68 billion of revenue in June, up 67.9% year over year and a new all-time monthly record, six percent above May, which was itself the previous record. Q2 came in at T$1.27 trillion (~$39.62 billion), 36% up year over year and above the top end of TSMC's own guidance. First-half 2026 revenue is now T$2.04 trillion, up 35.6% year over year. Full quarterly earnings land Thursday. The number that matters is this: every hyperscaler capex slide about AI datacenters has to actually show up as chip orders in Hsinchu, and it is showing up.

Google Cloud blog / Google DeepMind, July 9 2026 googlegoogle-cloud

Google put AlphaEvolve on Cloud, and the tell is that you have to write the scoring function yourself

AlphaEvolve, DeepMind's Gemini-based evolutionary code-optimization agent, went generally available on Gemini Enterprise Agent Platform on July 9, out of a private preview that ran from December 2025. Early customers include Klarna (doubled AI throughput, explored 6,000 candidate programs over three weeks), BASF (80%-plus improvement in planning models), Kinaxis (22% better forecasting accuracy, 90%-plus runtime reduction), JetBrains (15 to 20% IDE performance), Schrödinger (4x molecular discovery speedup), FM Logistic (10.4% better warehouse routing, 15,000 kilometers saved). Pricing not disclosed. AlphaEvolve is not a coding assistant. It is a search engine over algorithms, and the mandatory client-side evaluator is the whole product.

Monday, Jul 13, 2026

Reuters / Data Center Dynamics / TechDogs, July 10-14 2026 metairis-chip

Meta's first real in-house AI chip goes into production at TSMC in September, one week after Meta started charging for tokens

Meta plans to begin mass production of its Broadcom-designed, TSMC-fabricated 'Iris' accelerator in September, part of a stack meant to double datacenter compute from 7 gigawatts this year to 14 gigawatts in 2027. The chip cleared bug testing in six weeks. Full-year capex is projected at $125-145 billion, most of it aimed at AI infrastructure. The signal, coming days after Muse Spark 1.1 launched with a public pricing page, is that Meta is done being a customer of the AI supply chain. It is trying to be the supply chain.

Google Cloud blog / Bain & Company / The Next Web, April 9-July 14 2026 googlegoogle-cloud

Google's Cloud Next pitch is that the agent isn't the product, the platform for managing 200 of them is

At Cloud Next '26, Google renamed Vertex AI to the Gemini Enterprise Agent Platform and added eighth-generation TPUs, Model Garden access to 200-plus models (Gemini 3.1 Pro, Gemini 3.1 Flash Image, Lyria 3, Gemma 4, and third-party frontier weights), an agent-to-agent A2A protocol, Workspace Studio, and a governance layer aimed at IT departments who are already tired of shadow-agent proliferation. The pitch is not that Google has the best single agent. The pitch is that every large enterprise is about to run hundreds of them, and someone has to be the control plane.

Reuters / Cryptopolitan / Yahoo Finance, July 3-4 2026 unitreeunitree-robotics

China's biggest humanoid maker got cleared for a $619M IPO in 73 days, which is not the timeline a normal capital market runs at

Unitree Robotics got final CSRC approval on July 3 for a Shanghai STAR Market listing that will raise about 4.2 billion yuan ($619M) against an implied $6.18 billion valuation. The company reported 1.699 billion yuan in 2025 revenue and 278 million yuan in net profit, and the regulator moved the filing from submission to approval in 73 days, one of the fastest STAR reviews on record. American humanoid maker Agility went public via SPAC last week. Different capital markets, same underlying claim: humanoid revenue has crossed the line where public underwriting starts to make sense.

Sunday, Jul 12, 2026

Anthropic / Forbes / The NonProfit Times, June 15 - July 12 2026 anthropicclaude-corps

Anthropic is paying 100 twenty-somethings $85K each to move into a nonprofit for a year and figure out what AI is actually useful for

Applications for the first Claude Corps cohort close Friday. The program is a $150 million commitment from Anthropic to place 1,000 early-career fellows inside US nonprofits, government agencies, and public-interest orgs across three cohorts starting October 2026. Fellows get $85,000 in salary, benefits, a CodePath mentor, office hours with Anthropic engineers, and a Claude token budget. Host orgs get, in theory, someone who will actually build the internal AI system that every executive director has been asked about at every board meeting since 2023. It is a weird shape for a frontier lab. It is also, quietly, one of the more coherent bets any of them has made on where the next thousand deployments actually happen.

Meta AI blog / Fortune / DataCamp / BigGo, July 9-10 2026 metamuse-spark

Meta shipped a closed frontier model with a paid meter on it, three years after building its brand on the opposite of that

Meta Superintelligence Labs released Muse Spark 1.1 last Thursday: an agentic, multimodal, million-token-context model with computer-use, multi-agent orchestration, and a public preview of the Meta Model API. Pricing is $1.25 per million input tokens and $4.25 per million output tokens, aggressively below GPT-5.6 and Opus 4.8. The technical release notes are impressive. The strategic release notes are the whole story: Meta trained a closed frontier model, put a meter on it, and joined the market it spent three years undercutting with free weights. Alexandr Wang is running this playbook now, and the playbook is the one Meta swore it would never run.

Mistral AI blog / AI Weekly / TechJack Solutions, July 2-8 2026 mistralleanstral

Mistral's new proof-generating model spent its first ten days finding real bugs in open source, which is exactly the wrong kind of good news

Leanstral 1.5, Mistral's Apache-2.0-licensed formal verification model for Lean 4, shipped July 2 and is free during beta. It has 119 billion total parameters with 6 billion active per token, beats Claude Sonnet by 8 points at pass@16 on the FLTEval proof benchmark while costing 15x less, and, according to Mistral's own blog post, has already uncovered previously unknown bugs in open-source codebases. The technical achievement is real. The uncomfortable subtext is that if a mid-sized French AI lab can point a proof agent at popular open-source libraries and immediately surface undiscovered bugs, that says more about the software we all depend on than it does about the model.

Saturday, Jul 11, 2026

Courthouse News / Cooley / GeekWire / PYMNTS, March to July 2026 perplexitycomet

The Ninth Circuit has been sitting on the Amazon vs. Perplexity ruling for a month, and every silent Friday is a warning shot for agentic commerce

Oral arguments in Perplexity's appeal of the Amazon injunction wrapped on June 11 in Seattle. It is now July 12 and the panel has not filed. Meanwhile Comet keeps shipping, Buy with Pro keeps taking PayPal, and every other frontier lab keeps pitching investors on the same agentic-shopping graph. If the Ninth Circuit affirms, the entire product category has a Computer Fraud and Abuse Act problem it cannot code around. If it reverses, the biggest retailer on the internet has to figure out what a robot customer is, in public, with no rulebook.

Friday, Jul 10, 2026

Axios / TheNextWeb / AI Insiders, July 8-10 2026 openaiopenai-deployment-company

OpenAI's Deployment Company just bought its second Palantir clone in eight weeks, and the shape of AI money is officially not the model anymore

The OpenAI Deployment Company, launched in May 2026 with $4 billion earmarked for acquisitions, announced its second buy on July 8: Northslope, an applied-AI outfit founded by ex-Palantir engineers, joins Tomoro on the shelf. Terms undisclosed. What OpenAI is actually buying is a bench of hundreds of forward-deployed engineers, the Palantir-invented species that sits inside a customer's office and builds the AI system around the customer's actual workflow. The pitch under the pitch: model capability is a commodity, implementation is the moat, and OpenAI would like to be the consulting firm now.

Thursday, Jul 9, 2026

Reuters via Bloomberg / Taipei Times / Seeking Alpha, July 7-8 2026 deepseekinference-chip

DeepSeek is designing its own inference chip, joining every other AI lab that has now announced this before actually shipping one

Reuters reported Monday that Hangzhou-based DeepSeek is in the early stages of developing an in-house AI chip aimed at inference workloads, quietly hiring chip-design engineers without public postings, and holding discussions with foundry, chip-design, and memory partners. The goal is to reduce dependence on Nvidia and Huawei, both of which currently sit under its stack. Analyst Richard Windsor gave the honest ceiling: 'DeepSeek has almost no chance of selling silicon outside of China unless it gets access to leading-edge manufacturing.' Nvidia shares slipped about 1.5 percent in premarket trading the morning of the report.

TechCrunch / SiliconANGLE / Simon Willison, July 8 2026 openaigpt-live

OpenAI's new voice model will say 'mhmm' back at you while you talk, which is either endearing or the actual future of human-machine relations

OpenAI launched GPT-Live on July 8, a pair of full-duplex voice models (GPT-Live-1 in paid ChatGPT tiers, GPT-Live-1 mini free everywhere) that can listen and speak at the same time, backchannel with 'mhmm' and 'yeah' while you're talking, sit quietly when you're thinking, and quietly delegate any question that needs real reasoning to GPT-5.5 in the background. Product lead Atty Eleti pitched it as step one toward voice as a primary computing interface. Simon Willison, testing the preview, reported the model was laughing at things he had not intended as jokes.

Axios / Analytics Insight / Technobezz, July 8-9 2026 xaispacexai

Musk dropped Grok 4.5 four and a half hours before OpenAI's GPT-5.6 broad rollout, because the release note is the vibes

SpaceXAI launched Grok 4.5 to the public on July 9, roughly four and a half hours before OpenAI's broad GPT-5.6 release the same day. The model runs on xAI's new V9 foundation at about 1.5 trillion parameters, roughly three times the size of the v8-small architecture behind Grok 4.3. Supplemental training included data from Cursor, the AI coding startup SpaceXAI recently acquired. Musk called it 'an Opus-class model, but faster, more token-efficient and lower cost.' No system card. No third-party benchmarks. Not yet in the EU.

Wednesday, Jul 8, 2026

CX Today / Forbes, July 7-8 2026 microsoftcopilot

Microsoft shipped its enterprise sales and service agents to general availability on Monday, and by Wednesday Forbes had shown they can't remember what you told them last week

Microsoft on July 7 moved Sales Agent and Service Agent to general availability inside Dynamics 365 and the Microsoft 365 Copilot stack, with Sandvik Coromant and Northern Trust named as launch customers. Both agents run on Model Context Protocol connections into live CRM data and slot into Outlook, Teams, and Dynamics workflows. The next morning Forbes published a comparative test of Microsoft Copilot Cowork, Amazon Quick, and Claude Cowork on multi-step enterprise scenarios. Microsoft won the workplace-context tasks. Amazon won the action-chain tasks. All three shared the same fundamental failure mode: they treat every session as a clean slate.

Axios / CNBC / TNW, July 7-8 2026 openaigpt-5-6

OpenAI got the federal sign-off to broadly release GPT-5.6, and the White House immediately said it did not sign off on anything

Axios reported Tuesday that the Commerce Department's Center for AI Standards and Innovation completed its review of GPT-5.6 and cleared OpenAI to widen access beyond the 20-partner preview list it had been restricted to since June. Three tiers ship: Sol at the top with a max reasoning mode, Terra for cost-sensitive enterprise workloads, Luna for high-volume speed. Within hours the White House told CNBC it had given no 'green light, approval or clearance' and that release decisions 'rest entirely with the companies.' OpenAI says government access review should not be the long-term default. The general availability date is still not on the calendar.

Bloomberg / Electrek, July 7 2026 umanorthstar

The scientist who built Optimus's brain is now trying to build a rival in Paris, and Yann LeCun cut one of the first checks

Bloomberg reported Monday that Rémi Cadène, the former Tesla Autopilot researcher who ran the AI systems inside Optimus and then went to Hugging Face to build LeRobot, has come out of stealth with UMA. The Paris startup showed a lightweight humanoid prototype called Northstar Version 0, built by a nine-person team in nine months. UMA closed a $40 million seed from Greycroft, Relentless, and Unity Growth, with angel checks from Yann LeCun, Datadog CEO Olivier Pomel, and Hugging Face co-founder Thomas Wolf. Cadène is talking to 50 potential customers about logistics, manufacturing, and healthcare pilots this year. The bet is that Europe's aging workforce and dense industrial base are the actual first market for humanoids, not California.

Tuesday, Jul 7, 2026

SiliconANGLE, July 6 2026 anthropicterawulf

Anthropic just signed a twenty-year $19 billion lease on a former Kentucky aluminum smelter, and the transmission wires are the reason

TeraWulf announced Monday it will build a 401 megawatt AI campus for Anthropic on a 790-acre former aluminum smelting site in Hawesville, Kentucky. Twenty-year term, roughly $19 billion in contracted revenue, initial capacity in the second half of 2027, full by early 2028. On the same day, TeraWulf sold its 168 megawatt Abernathy, Texas facility to Fluidstack for $450 million, and Anthropic has been in a Fluidstack partnership since November 2025. This is a lab that is no longer procuring like a lab.

Forbes, July 2 2026 boston-dynamicsatlas

The new Atlas is about 9 times simpler than the old one, and Hyundai is standing up a line to build 30,000 of them a year

Boston Dynamics director of robot behavior Alberto Rodriguez told Forbes on July 2 that the fifth-generation Atlas is 'almost an order of magnitude reduction in complexity' versus the prior version. Fewer parts, less unique tooling, faster assembly, lower cost. Hyundai, which owns Boston Dynamics, plans 30,000 units a year. The previous Atlas ran north of $200,000 per unit. The 2026 build is reserved for Hyundai's own factories. External sales come later. This is what humanoid mass production actually looks like when the design brief is 'now build 30,000 of them.'

SemiAnalysis / Tom's Hardware / CNBC, July 6 2026 nvidiakyber

The one motherboard Nvidia cannot yet build reliably just pushed its next AI rack to 2028, and it took every PCB stock in Asia down with it

SemiAnalysis reported Monday that Nvidia's Kyber NVL144 rack, showcased by Jensen Huang at GTC three months ago, has slipped more than twelve months and will not ship until 2028. The problem is a 78-layer orthogonal PCB midplane that has to interconnect 144 GPUs at SerDes speeds above 448G. Nvidia's supply chain cannot yet manufacture it at yield. Ibiden fell 10 percent. Elite Material fell 10. Samsung Electro-Mechanics fell 11. Kingboard Laminates fell 18. Rubin Ultra was trimmed from four dies to two. Nvidia told Bloomberg the roadmap is 'intact.'

Monday, Jul 6, 2026

TechCrunch, July 5 2026 agility-roboticsdigit

Agility Robotics is going public via SPAC at $2.5 billion, and its CEO would like to remind you a Digit is not coming to your kitchen

Agility Robotics announced a merger with Churchill Capital Corp XI on July 4, valuing the company at about $2.5 billion and raising more than $620 million in gross proceeds, including a $200 million PIPE led by Foxconn. The combined entity will trade on Nasdaq as AGLT before year-end. CEO Peggy Johnson used the press tour to spike, on live camera, every question about home robots. Warehouses are structured. Homes have dogs and babies. The company would prefer investors keep that distinction firmly in mind.

Fortune, July 5 2026 boston-dynamicsatlas

A Boston Dynamics Atlas robot did an Erling Haaland goal celebration at the World Cup and did not fall over, which is somehow the real news

During halftime of the Brazil-Norway Round of 16 match on July 5, at New York/New Jersey Stadium in front of 80,000 people, Hyundai wheeled out a fifth-generation Atlas that walked to midfield, performed goal celebrations in the style of Haaland, Harry Kane, Matheus Cunha, and Son Heung-min, then handed the match ball to the referee. Nothing broke. The interesting part is not the celebrations. The interesting part is that Boston Dynamics ran a humanoid live, on natural grass, in front of a global TV audience, without a failure state that made the highlight reels.

TechCrunch, July 2 2026 microsoftfrontier-company

Microsoft is spinning up a 6,000-person AI consulting arm because nobody is actually deploying the AI they paid for

Microsoft announced on July 2 a new operating unit called Microsoft Frontier Company, funded at $2.5 billion and staffed with 6,000 engineers, whose job is to physically go inside London Stock Exchange Group, Unilever, Land O'Lakes, and Accenture and get their Microsoft AI tools past the pilot stage. It lands two days after Amazon committed $1 billion to the same thing, following earlier launches from OpenAI and Anthropic. The forward-deployed engineer model has stopped being a Palantir joke and started being the default enterprise-AI playbook. This is what the 2026 procurement cycle actually looks like.

Sunday, Jul 5, 2026

TechCrunch, July 2 2026 anthropicsamsung

Anthropic is talking to Samsung about a custom chip, and the last independent frontier lab just joined the silicon race

TechCrunch reported on July 2 that Anthropic is in early discussions with Samsung about a custom chip. Details are thin: no timeline, no budget, no confirmed purpose. The announcement lands one week after OpenAI unveiled its Broadcom-designed Jalapeño inference chip. Anthropic's statement is a masterpiece of noncommittal: a diversified hardware stack including Google, Amazon, and Nvidia chips will remain pivotal, and on Samsung the company had 'nothing further to add.' The point is that they are talking at all.

Anthropic newsroom, June 30 2026 anthropicclaude-science

Anthropic launched a drug-discovery workbench with sixty scientific tools built in, and Novo Nordisk was already in the room

Claude Science shipped in beta on June 30 for Pro, Max, Team, and Enterprise. It bundles more than sixty curated skills across genomics, proteomics, and cheminformatics, plus a set of reviewer agents that check citations and calculations. Anthropic also announced its own internal drug-discovery program, aimed at neglected diseases the big pharmas skip. Early customers include Novo Nordisk, the Allen Institute, and UCSF. It is the first frontier lab product that reads less like a chat product and more like the Bloomberg terminal for a wet lab.

Teslarati, July 1 2026 teslaoptimus

Tesla's Optimus V3 production line is walking around Fremont, and Musk says the first robots will be extremely slow

On July 1 Elon Musk toured what he calls the first Optimus production line, standing in space that until May was building the Model S and Model X. First units come off the line in late July or August. The stated Fremont ceiling is one million robots per year. Musk's own framing was the least VC-tweet-shaped thing he has said in months: 'Optimus production will be extremely slow at first. This is not like making a car.' Each robot has around ten thousand unique parts.

Saturday, Jul 4, 2026

Tom's Hardware, July 3 2026 openaisam-altman

OpenAI is offering the US government a $42 billion equity stake in the same week Anthropic crossed it on revenue, which is a very specific kind of pitch to be making right now

Sam Altman floated a 5 percent OpenAI stake to Washington on July 3, six days after the government asked OpenAI to slow-roll GPT-5.6 to an approved-customer list. At the $852 billion March valuation, that stake is worth about $42.6 billion. In the same week, Anthropic passed OpenAI on annualized revenue, roughly $30 billion to $25 billion. The GPT-5.6 delay and the equity offer look less like a policy proposal and more like a company negotiating from a weaker position than it had 90 days ago.

Friday, Jul 3, 2026

llm-stats July 2026 model updates openaigpt-5-6

OpenAI is running the GPT-5.6 family through a 20-org US government access list, which is a very different pitch than the mission statement

OpenAI previewed the GPT-5.6 family (codenames Sol, Terra, Luna) on June 26 and gated access to roughly 20 organizations on a US-government approved list. It is the first frontier release in years that a Free-tier user cannot try at any price. Meanwhile Anthropic just made Claude Sonnet 5 the default model on every Claude account at $2 per million input tokens. The two positioning bets could not be further apart. One reads as safety infrastructure, one reads as a growth engine, and both companies are pricing them like they mean it.

Thursday, Jul 2, 2026

TechCrunch metameta-compute

Meta wants to sell you AI compute now, and CoreWeave's stock heard every word of it

Bloomberg reported on July 1 that Meta is building a cloud infrastructure business called Meta Compute, led by infrastructure head Santosh Janardhan, Superintelligence Labs leader Daniel Gross, and president Dina Powell McCormick. It will rent raw GPU capacity in the CoreWeave style, and let developers hit hosted Meta models the AWS Bedrock way, including the newly closed-weight Muse Spark. Meta shares jumped roughly 9 percent on the news. CoreWeave closed down about 14 percent. The context is a $182.9 billion AI infrastructure commitment that Meta has been struggling to justify with its own consumer AI revenue, so now the excess capacity becomes the product.

Tech Xplore humanoid-robotseurope-robotics

The Netherlands opened a humanoid robotics center to catch a race China is already 85% of

The Humanoid Application Center opened near Rotterdam on July 2, positioned by CEO Evert Jaap Lugt as Europe's play to close a gap it has already lost most of. Barclays data cited by the center puts China at 85 percent of the world's humanoid installations last year. The pitch is a bridge between European industrial buyers and humanoid vendors, and the first live deployment is already lined up: real estate developer Niels Langenhuizen plans to put a humanoid on a building site by year's end, to help hit the Dutch government's target of 100,000 new homes a year. Lugt's projection: within five years, humanoids will be visually indistinguishable from humans at five meters. That is either a startup pitch or a horror movie logline.

The Register ubtechhumanoid-robots

UBTECH's new humanoid has 88 joints, 20ms lip sync, and the world's first 'emotion-aware LLM,' which is a phrase somebody typed

Chinese robotics maker UBTECH announced the UWORLD U1 on July 2, marketed as the world's first full-size mass-produced ultra-bionic humanoid robot. The pitch: 88 degrees of freedom, 90 percent of fundamental human movements, a dual-pivot biomimetic cervical spine, and a 20-millisecond speech-to-lip-sync latency, all wrapped around what UBTECH is calling the world's first emotion-aware LLM, said to recognize over 20 emotional states at 90 percent accuracy. Price is $17,600. Orders reportedly already past 13,000. The Register's coverage compared them to creepy pop-star action figures. That is not entirely wrong.

Wednesday, Jul 1, 2026

Anthropic anthropicclaude

Anthropic shipped Claude Sonnet 5 at $2 in, $10 out, which is the sound of the price race finally hitting the agent tier

Anthropic released Claude Sonnet 5 on June 30 with introductory API pricing of $2 per million input tokens and $10 per million output tokens through August 31, after which it moves to $3 and $15. It is the default model on Free and Pro, available on Max, Team, and Enterprise, and lives inside Claude Code. The pitch is 'the most agentic Sonnet yet' with stronger browser, terminal, and long-horizon tool use. The pricing is the actual news. Anthropic is pointing a very cheap, very autonomous model directly at the agentic coding workload that GitHub Copilot just started billing at API rates last month.

Forbes xaigrok

Grok is now mostly a porn app, and its traffic is falling anyway

A report from The Information, picked up by Forbes and Engadget last week, finds that adult content accounts for well over half of Grok's traffic, and that xAI is now actively investing engineering resources into its explicit image and video generation as a growth strategy. In the same window, Grok's overall web traffic fell 22% between January and May 2026, more than any other major chatbot. Claude's traffic in the same window grew 369%. Google's Gemini grew 40%. Grok is winning a race the rest of the field decided not to enter, in a category that appears to be shrinking under it.

VentureBeat googlegemini

Google shipped an enterprise image model called Nano Banana 2 Lite, and yes, that is officially its name

On June 30 Google rolled out Gemini 3.1 Flash-Lite Image, publicly nicknamed Nano Banana 2 Lite, for $0.034 per 1,000 images at roughly four-second generation latency, targeted at enterprise catalog and product-image use cases. It hit Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform on day one. Alongside it, Gemini Omni Flash entered public preview as a natively multimodal video generation and editing model. The naming remains fully committed to the bit. The pricing quietly moves image generation into the same commodity utility zone that transcription and embeddings landed in eighteen months ago.

Tuesday, Jun 30, 2026

TechCrunch 8090-labschamath-palihapitiya

Chamath Palihapitiya is running a company again, and Salesforce Ventures is paying $135 million to watch him do it

8090 Labs closed a $135 million Series A on June 29 led by Salesforce Ventures. Chamath Palihapitiya, who founded the enterprise AI coding company in January 2024, is stepping in as full-time CEO. It is his first operating role since leaving Facebook in 2011. The cap table reads like a screenshot of the All-In podcast subscriber list: David Sacks, Jason Calacanis, David Friedberg, Jeffrey Katzenberg, plus angels Nikesh Arora and Adam D'Angelo. The strategic twist is Salesforce leading. The company whose entire business is packaged software just funded a shop whose entire pitch is that software gets built, not bought.

Wilson Sonsini coloradocolorado-ai-act

Colorado's landmark AI law was supposed to take effect yesterday, and instead it got quietly gutted six weeks ago

When Governor Jared Polis signed SB 205 in 2024, Colorado became the first US state with a comprehensive high-risk AI law. It was scheduled to take effect on June 30, 2026, with a $1 million per violation penalty for algorithmic discrimination in employment, healthcare, housing, finance, insurance, and legal services. On May 14, 2026, Polis signed SB 189, which repealed and replaced SB 205 before it ever went live. The new version delays the effective date to January 1, 2027, strips the duty of care against algorithmic discrimination, and eliminates the risk management program requirements. What is left is a transparency-and-disclosure regime with the enforcement teeth pulled.

Al Jazeera south-koreasamsung

South Korea just pledged a trillion dollars to chips and AI data centers, which is the sort of number that used to be a typo

On June 29 President Lee Jae Myung announced a national AI and semiconductor buildout worth more than 1,000 trillion won, roughly $1 trillion. Samsung and SK Hynix together commit 800 trillion won for new fabs in the southwest. SK Group, GS Group, and Naver commit 550 trillion won for AI data centers targeting 18.4 gigawatts of capacity by 2035. A separate 81 trillion won packaging cluster gets added near Seoul. The framing is a 'triple axis' of semiconductors, physical AI, and data centers. The tell is that the total is bigger than the annual GDP of Poland, Taiwan, or Saudi Arabia.

Monday, Jun 29, 2026

SiliconANGLE china-roboticshumanoid-robots

Two Chinese humanoid startups hit $2.8 billion valuations on the same Monday, which is its own data point

AI2 Robotics and X Square Robot both crossed $2.8 billion in fresh funding rounds announced on June 29. AI2 raised roughly $735 million for its AlphaBot wheeled-humanoid platform. X Square closed four consecutive rounds into a Series C, with lead checks from Alibaba, ByteDance, Meituan, and Xiaomi, making it the only embodied-AI company in China backed by all four internet giants. Coming six days after Morgan Stanley raised its 2026 China humanoid shipment forecast to 50,000 units, the dual milestone reads like the capital side of the same curve catching up.

Data Center Knowledge qualcommmeta

Meta is buying Qualcomm CPUs for its data centers, which is the strangest sentence in semiconductors this week

At its June 24 Investor Day, Qualcomm announced a multigenerational CPU deal with Meta, a second unnamed hyperscale customer, the Dragonfly product family for AI data centers, and a new FY29 target of $40 billion in non-handset chip revenue, nearly double the previous $22 billion guidance. The data center business inside that target is pegged at over $15 billion a year. Qualcomm has spent a decade as the company that puts modems in phones. As of last Wednesday, it is now also a company whose chips Meta has agreed to buy for its server fleet.

TechTimes github-copilotmicrosoft

GitHub Copilot's first usage-based bills are landing today, and the screenshots are starting to circulate

GitHub Copilot switched from flat-rate to token-metered billing on June 1. June 30 closes the first full month under the new regime. Screenshots are circulating of developers whose bills jumped from $29 to roughly $750, and from $50 to roughly $3,000. The phrase doing the rounds in the GitHub community forum is 'what a joke.' GitHub's position is that the old economics were unsustainable. The market's position is that nobody costed in their pair programmer running on a taxi meter.

Sunday, Jun 28, 2026

CNBC anthropicalibaba

Anthropic told the Senate that Alibaba ran 25,000 fake accounts at Claude to clone the parts they liked

A June 10 letter from Anthropic to the Senate Banking Committee, surfaced this week, alleges that operators tied to Alibaba's Qwen lab put 28.8 million queries through Claude between April 22 and June 5 using 25,000 fraudulent accounts. They were specifically pulling on coding and agent reasoning. Two senators are now writing a sanctions amendment into the defense bill.

TechCrunch openaigpt-5-6

OpenAI shipped GPT-5.6 Sol to twenty companies and the rest of you have to ask the government for permission

OpenAI previewed GPT-5.6 Sol, Terra, and Luna on Friday with state-of-the-art numbers on Terminal-Bench and cybersecurity evals. Then it announced that only about twenty companies can use them, because the Trump administration asked OpenAI to gate the release. It is the first time a US frontier-model launch has shipped with a federally vetted access list attached.

CNBC humanoid-robotschina-robotics

Morgan Stanley doubled its China humanoid forecast to 50,000 units, and they have already doubled it once this year

Morgan Stanley raised its 2026 forecast for China humanoid robot shipments from 28,000 to 50,000 units on Tuesday. The original January number was 14,000. The bank cites accelerating commercial validation, a 6.8 billion yuan State Grid order, and a supply chain that has stopped being the bottleneck. By 2030 the market is projected at $15 billion and 446,000 units a year.

Friday, Jun 26, 2026

Bloomberg openaianthropic

OpenAI is now planning to IPO after Anthropic. The smaller company gets the headline first because the bigger one wants a number Wall Street is not yet ready to print.

Bloomberg reported Friday that OpenAI is now eyeing a 2027 listing and expects Anthropic to take the public market first. Both companies have already filed confidential S-1s. Anthropic filed June 1 targeting an October 2026 Nasdaq debut at a $965 billion valuation. OpenAI was targeting a fall 2026 window. That window is now reportedly closed.

Thursday, Jun 25, 2026

Figure figure-aifigure-03

Figure says its BotQ factory now builds one humanoid robot per hour. That is 24x faster than four months ago and still slower than a Honda CR-V.

Figure reported this week that its BotQ manufacturing facility has reached a one-robot-per-hour cycle time for the Figure 03 humanoid, a 24x throughput improvement over the previous 120 days. The company has shipped more than 350 units, targets 12,000 robots in year one of the first line, and is aiming for 100,000 total over four years.

Bloomberg googledeepmind

Four senior AI researchers left Google in six days. Two went to Anthropic, one went to OpenAI, and the through-line is the IPO calendar.

Bloomberg reported June 24 that Jonas Adler and Alexander Pritzel, both flagged internally as key Gemini contributors, are leaving Google for Anthropic. Their departures follow Nobel laureate John Jumper announcing the same move on June 20, and Noam Shazeer's defection to OpenAI a week earlier. Four senior departures, two destinations, one timing window.

OpenAI openaibroadcom

OpenAI and Broadcom shipped Jalapeño, OpenAI's first custom chip. The model that wrote half of it was trained on the GPUs it is designed to replace.

OpenAI and Broadcom unveiled Jalapeño on June 24, the first accelerator in a multi-generation co-design partnership. Engineering samples are already running GPT-5.3-Codex-Spark in the lab at production frequency, the design cycle was nine months from sketch to tape-out, and OpenAI claims roughly 50 percent cost savings versus typical AI GPUs.

Thursday, Jun 18, 2026

Anthropic anthropicseoul

Anthropic opened a Seoul office two days after Fable 5 got export-banned. The customer list announced with it is the actual move.

Anthropic confirmed its Seoul office on June 17 with KiYoung Choi as Representative Director, plus enterprise deployments at NAVER, Samsung SDS, LG CNS, Nexon, Hanwha Solutions, and Channel Corp. It is the third Anthropic Asia-Pacific office after Tokyo and Bengaluru, and the timing relative to the Commerce Department's Fable 5 ruling is not subtle.

NEJM AI openaio3

OpenAI's o3 Deep Research solved 18 rare childhood disease cases that human specialists had already given up on. The hit rate is the actual story.

A Boston Children's Hospital and OpenAI study published in NEJM AI on June 18 ran 376 previously unsolved rare-disease cases through the o3 Deep Research reasoning model and produced clinically confirmed diagnoses in 18 of them, a 4.8 percent additional yield on top of expert analysis that had stalled. The cases were on shelves for years.

Tech Times teslaoptimus

Tesla is turning the Fremont Model S line into the Optimus Gen 3 line. The Model S was the car that saved Tesla. The robot is what it gets to be when it grows up.

Heading into next week's shareholder meeting, Tesla has confirmed that the Fremont production line that built the Model S will be converted to assemble Optimus Gen 3, with low-volume production targeted for late 2026 and a stated long-term capacity of one million robots per year out of a dedicated Texas plant. Internal Fremont deployments are already running.

Wednesday, Jun 17, 2026

About Amazon amazonwarehouse-robotics

Amazon's one millionth warehouse robot shipped to a fulfillment center in Japan. The new AI dispatching the fleet was trained on the previous 999,999.

Amazon announced this month that its warehouse robot fleet has crossed one million units across more than 300 fulfillment centers. The same announcement introduced DeepFleet, a foundation-model-style fleet coordination system trained on Amazon's historical robot-movement data and reportedly improving travel efficiency by 10 percent across the network.

TechCrunch spacexxai

SpaceX is acquiring Cursor for $60 billion in all-stock, which means a rocket company now owns the most-used AI coding tool, and that sentence is allowed to exist.

SpaceX confirmed Tuesday it will buy Anysphere, the company behind Cursor, in a $60 billion all-stock deal that triggers off a buy-or-pay-$10-billion option set up before SpaceX's IPO in April. SpaceX stock closed up about 17 percent and briefly passed Microsoft in market cap. The Anysphere founders did not sell a share and roughly doubled their paper net worth.

Tuesday, Jun 16, 2026

MarkTechPost openaideployment-simulation

OpenAI now tests new models by feeding them 1.3 million old conversations and watching what changes. The error bar is the actual story.

OpenAI introduced Deployment Simulation on Monday, a pre-release method that strips assistant responses from anonymized historical conversations, regenerates them with a candidate model, and scores the new outputs for safety violations. The paper reports a median 1.5x multiplicative error on predicted failure rates. The tail error is closer to 10x.

Monday, Jun 15, 2026

Sunday, Jun 14, 2026

Cursor cursorcursor-2

Cursor 2.0 ships, and the editor is finally trying to be a team orchestration layer instead of a fancy autocomplete

Cursor 2.0 dropped over the weekend, six weeks after the Windsurf merger closed. The headline feature is multi-agent team workflows: long-running background agents that file PRs against a shared repo, with a routing layer that lets you assign different tasks to different models. The pitch is less 'AI inside your editor' and more 'AI is your editor now.'

Friday, Jun 12, 2026

Anthropic pricing and model card anthropicclaude

RAG is not dead. It just got a smaller job.

A month after Anthropic shipped Claude Opus 4.7 with a 1M-token context window, dev teams are quietly retiring their bespoke retrieval pipelines for the under-800-page case. The interesting question is what RAG keeps now that 'just read the thing' is a viable answer.

Thursday, Jun 11, 2026

Mercedes-Benz apptronikapollo

Mercedes just put Apptronik's Apollo on the Tuscaloosa SUV line

Mercedes-Benz and Apptronik announced Thursday that the Apollo humanoid will deploy to the Tuscaloosa, Alabama plant in early 2027 for parts-kitting and pre-assembly work. Mercedes is already an Apptronik Series A and Series B investor, which is why the deal skipped a competitive bake-off. With Figure at BMW Spartanburg, Atlas at Hyundai Georgia, and now Apollo at Mercedes Tuscaloosa, the German-and-Korean OEMs have locked in three of the four humanoid production lines that exist.

Reuters samsungsk-hynix

Samsung finally got its HBM4 past Nvidia's qualification cycle

Samsung Electronics confirmed Friday that its 12-high HBM4 modules cleared Nvidia's qualification audit for Vera Rubin Ultra accelerators, ending roughly eighteen months of failed cycles. Samsung will supply Vera Rubin Ultra production volumes starting Q4 alongside SK hynix and Micron. SK hynix shares dropped 4.2% in Seoul trading. Samsung jumped 6.8%. The Korean memory duopoly is back to being a duopoly.

Snowflake snowflakesnowflake-summit

Snowflake's pitch to the enterprise: keep the data, rent the brain

Snowflake announced general availability of Cortex Agents on the final day of Snowflake Summit, letting customers orchestrate Claude Fable 5, Gemini 3 Pro, GPT-5.5, and Snowflake's own Arctic models against data already sitting inside the warehouse. No egress, no copy, governance handled by existing Snowflake row-level security. The shot is aimed squarely at Databricks AI/BI and reframes the model-routing wars as a question of where the data lives.

Wednesday, Jun 10, 2026

Apple Developer applefoundation-models

Apple's Foundation Models framework treats Claude and Gemini as plugins

Apple's WWDC 2026 Foundation Models update introduced a public Swift LanguageModel protocol that third-party providers implement to slot into iOS apps. Anthropic's Claude and Google's Gemini conform on day one. Apple's own on-device model conforms too. The developer writes session logic once, and swaps providers by changing a single import line. Apple just made the model the part of the stack it does not need to win.

European Commission eu-ai-actarticle-50

Brussels finalized its deepfake labeling rules. The watermarks that actually work were not included.

The European Commission published the final Code of Practice for AI-generated content on Wednesday, ahead of the Article 50 transparency obligations under the AI Act taking effect August 2. Deepfakes (defined broadly) and AI-generated public-interest text must be visibly marked. Systems on the EU market before August 2 get a transitional window through December 2. New launches after August 2 must ship compliant from day one.

OpenAI openaioracle

OpenAI moved into the Oracle procurement portal

OpenAI and Oracle announced Wednesday that Oracle Cloud Infrastructure customers can apply existing Oracle Universal Credits toward OpenAI models and Codex. Rollout begins in the coming weeks. It is the third hyperscaler distribution channel for OpenAI after Azure and AWS, and the cleanest path yet for an enterprise whose finance team already pre-committed billions to a non-Microsoft cloud.

Tuesday, Jun 9, 2026

Boston Dynamics boston-dynamicsatlas

Boston Dynamics has already sold every Atlas it will build this year

Every unit of the electric Atlas slated for 2026 production is committed, with the entire run going to Hyundai's Robotics Metaplant Application Center and Google DeepMind. Hyundai is targeting 30,000 units a year by 2028 across Kia and Hyundai assembly lines. DeepMind is integrating Gemini Robotics as Atlas's reasoning layer. The bottleneck for the rest of the humanoid market is now Boston Dynamics's own manufacturing line in Massachusetts.

Yahoo Finance / Reuters intelintel-foundry

Google ordered three million TPUs from Intel, which is now a sentence that makes sense

Intel closed up 11.2% Tuesday after reports that Alphabet placed an Intel Foundry order for more than three million Tensor Processing Units, scheduled for 2028 delivery. AMD added 5.1%, Broadcom 2.8%, and the XLK tech ETF closed up 2.15%. The read on the order: the largest non-NVIDIA AI silicon customer in the world is willing to sole-source a TPU generation through an American foundry, and that foundry is no longer TSMC.

Anthropic anthropicclaude

Anthropic put its Mythos model on a leash and shipped it to everybody

Claude Fable 5 launched Tuesday as the first publicly available Mythos-class model. Same underlying model, hard classifiers that bounce cybersecurity, biology, chemistry, and distillation prompts back to Opus 4.8. Priced at $10 in / $50 out per million tokens, twice Opus 4.8. Free inside Pro and Max plans through June 22, then credits-only. Available on the Anthropic API and AWS Bedrock day one.

Monday, Jun 8, 2026

Apple Newsroom applewwdc

Apple shipped Siri AI to a slide deck. The actual Siri AI ships 'later this year.'

WWDC 2026 announced Siri AI, branded it, gave it an app, ran the sizzle reel, then put it in beta 'later this year' on supported devices in English only. The rest of Apple Intelligence is in the developer beta as of Monday. The EU and China are excluded. It is Tim Cook's last WWDC, with John Ternus taking over September 1.

Korea JoongAng Daily nvidiajensen-huang

Jensen Huang spent five days in Korea and came back with six signed partnerships

NVIDIA's CEO swung through Seoul June 4-8 after Computex, met Hyundai, LG, SK, Samsung, and Naver, and left with six announced or signed deals across memory, gigawatt AI cloud, automotive, robotics, and a national R&D center. Korea is no longer just a memory supplier to NVIDIA. It is now a sovereign-AI customer, an OEM partner, and the location of a Seoul-based NVIDIA R&D facility.

Bloomberg uberwayve

Uber and Wayve quietly opened London's robotaxi waitlist, and Waymo is right behind them

Uber added a Wayve interest list to the London app on Monday. Launch is 'in the next couple of months,' contingent on UK regulators finalizing the framework. Fleet is Ford Mustang Mach-E EVs with a human safety driver in the early phase. Waymo is also entering London, setting up Europe's first head-to-head robotaxi market between two materially different stacks.

Sunday, Jun 7, 2026

Friday, Jun 5, 2026

AppleInsider: iOS 27, macOS 27, Siri preview applewwdc

WWDC opens Monday, and the Siri 2.0 test is whether it ships in the developer beta the same afternoon

The WWDC 2026 keynote is June 8 at 10am Pacific. The credibility test for Apple Intelligence is whether the rebuilt Siri lands in the iOS 27 developer beta that day, not 'rolling out throughout the year.' The leaks say a custom 1.2 trillion parameter Gemini model is doing the cloud reasoning under the hood for roughly $1 billion a year.

Thursday, Jun 4, 2026

Global Banking & Finance intelfoxconn

Intel found a friend at Computex, and his name is Foxconn

Foxconn and Intel announced a strategic collaboration on next-generation rack-scale AI infrastructure at Computex 2026, combining Xeon CPUs and Intel's AI accelerator line with Foxconn's manufacturing and systems integration. No timeline, no customers, no dollar figure.

Wednesday, Jun 3, 2026

Tuesday, Jun 2, 2026

Monday, Jun 1, 2026

Sunday, May 31, 2026

Friday, May 29, 2026

Thursday, May 28, 2026

TechCrunch cognitiondevin

Cognition raised a billion at $26B, doubled its valuation in eight months, and the only independent coding agent left standing is now priced like one

Devin's maker closed a $1B round led by Lux Capital, General Catalyst, and 8VC at a $26 billion post-money valuation, more than double its September mark. After Cursor merged with Windsurf and Karpathy migrated to Anthropic, Cognition is the last large independent in the coding-agent category, and the round is priced accordingly.

Wednesday, May 27, 2026

ERP Today servicenownvidia

ServiceNow and Nvidia announced an agent that runs on your desktop in a sandbox you can audit, which is a less fun pitch than 'agentic AI for everyone'

Project Arc, unveiled at Knowledge 2026, is a desktop agent built on ServiceNow Action Fabric, secured by Nvidia OpenShell, and governed through ServiceNow's AI Control Tower. The interesting part is not the agent. It is that the company finally chose to lead with the governance plumbing instead of the demo.

Tuesday, May 26, 2026

Tom's Hardware nvidiaasic

Custom AI ASICs are growing faster than Nvidia GPUs for the first time, and the hyperscalers built the wedge themselves

TrendForce projects 44.6 percent shipment growth for custom AI ASICs in 2026 versus 16.1 percent for merchant GPUs. Counterpoint says global ASIC shipments will triple from 2024 to 2027. Google TPU, AWS Trainium, Microsoft Maia, and Meta MTIA are all hitting full production at the same time, and their procurement budgets are quietly diverting away from the company that sold them last cycle's gear.

The Robot Report roboticshumanoids

The Robotics Summit opened in Boston this morning, and humanoids have quietly stopped being a demo

The 2026 Robotics Summit & Expo kicked off May 27 with a 'State of Humanoids' keynote panel and a set of milestones that are no longer projections. Agility's Digit just crossed 100,000 totes in live commercial work, Boston Dynamics has sold out its entire 2026 Atlas production run, and the room is starting to talk like a manufacturing industry instead of a research field.

Monday, May 25, 2026

Sunday, May 24, 2026

Friday, May 22, 2026

Thursday, May 21, 2026

Wednesday, May 20, 2026

Tuesday, May 19, 2026

Sunday, May 17, 2026

Saturday, May 16, 2026

Friday, May 15, 2026

Thursday, May 14, 2026

Tuesday, May 12, 2026

Monday, May 11, 2026

Sunday, May 10, 2026

Saturday, May 9, 2026

Friday, May 8, 2026

Wednesday, May 6, 2026

Tuesday, May 5, 2026

Sunday, May 3, 2026

Saturday, May 2, 2026

Friday, May 1, 2026

Monday, Apr 27, 2026

Sunday, Apr 19, 2026

Saturday, Apr 11, 2026