BriefTechNews
Anthropic's $81B IPO, a steering-wheel-less Model Y, and the day AI ate hardware

Anthropic's $81B IPO, a steering-wheel-less Model Y, and the day AI ate hardware

7 min read · 15 sources

TL;DR
  • Anthropic raised $81.3 billion at $83 per share, topping SpaceX's record IPO and valuing the company at $1.5 trillion.
  • Tesla will build an autonomous Model Y without a steering wheel or pedals, seating six or seven passengers.
  • AWS released ExtendDB 1.0, a DynamoDB-compatible adapter running on PostgreSQL, SQLite, Cassandra, and MongoDB.
  • Google's Gemini 4 Neon costs $3 per million input tokens and scores 65.4% on CursorBench 4.1, beating Claude Opus 5.5.
  • Nvidia shares are down 37% from their 52-week high as interim CEO Colette Kress told employees to move forward.

Anthropic just pulled off the largest IPO in history, and it wasn’t close. Priced at $83 per share - above the already-aggressive $70 - $78 range - the AI lab raised $81.3 billion by selling roughly 980 million shares. Fully diluted, that’s a $1.5 trillion market cap, and trading starts Thursday on the Nasdaq under “ANT.” For context, this tops SpaceX’s previous record, and it lands amid market jitters over interest rates and the leadership chaos at Nvidia. This is the moment the AI industry’s valuation story gets stress-tested in public markets.

The rest of the day is a mix of hardware without steering wheels, databases that pretend to be DynamoDB, and a physics paper that might make jamming impossible. Here’s what actually matters.

Anthropic priced at $83 per share, raising $81.3 billion and topping SpaceX’s record by a wide margin.

Anthropic's $81.3B IPO resets the AI valuation board

Source: cnbc.com ↗

The numbers are staggering. Anthropic priced its IPO at $83 per share, raising $81.3 billion and beating SpaceX’s record. The fully diluted valuation sits around $1.5 trillion. Amazon and Google each hold stakes now worth over $100 billion, and Google, Microsoft, Nvidia, and AMD collectively agreed to invest $26 billion at the IPO price.

For engineers, this isn’t just a finance story. It’s the first real test of whether the AI capex cycle can convert into public-market returns. OpenAI’s IPO is expected in 2027, which means Anthropic’s stock performance becomes the benchmark every other AI lab gets measured against. If ANT craters, the next 18 months of AI funding gets a lot harder.

Tesla's Model Y without a steering wheel

Source: insideevs.com ↗

Elon Musk announced at the NYT DealBook Summit that Tesla will build a fully autonomous Model Y with no steering wheel or pedals, designed to seat six or seven passengers. This follows the two-seat Cybercab, which covers about 90% of passenger miles and costs under $30,000. The new variant targets the remaining 10% that needs more seats.

Regular Model Ys with steering wheels stay in production. No timeline or pricing was given, which is typical. The interesting engineering note is the sequencing: Tesla is going from a purpose-built two-seater to a modified existing platform. That suggests the autonomy stack, not the vehicle hardware, is the constraint. If the stack is ready for the Cybercab, the Model Y variant is mostly a packaging problem.

Hardware is software, and software is eating aerospace

Source: unchartedterritories.tomaspueyo.com ↗

Boom Supersonic CEO Blake Scholl makes the case that hardware engineering is increasingly software engineering. Most calculations are done in spreadsheets rather than as robust software systems. He argues for designing algorithms for hardware design, not just the hardware itself, enabling rapid redesign under different assumptions.

The numbers are the point: AI now lets one engineer do in realtime what previously took 100 engineers six months, and one tool auto-generates 65% of an entire aircraft structure. For anyone who has watched a hardware team spend weeks re-running a simulation because someone changed a parameter, this reframing is overdue. The practical takeaway is that hard tech companies that treat their design process as software will out-iterate those that don’t.

Unblockable communication with gravity

Source: science.org ↗

Every wireless technology you use is electromagnetic, which means it can be screened, jammed, or blocked. A new study in Science Advances proposes using gravity instead. The protocol works by a broadcaster encoding a binary message through moving a mass, and a receiver detecting the gravitational signal with an off-the-shelf, hand-held gravimeter.

This is real, implemented, and slow. Gravity is weak, so data rates will be abysmal. But for scenarios where you need a channel that literally cannot be blocked - submarines, deep mines, or contested environments where RF jamming is a given - this is a fundamentally different threat model. The paper is worth reading if you care about comms resilience, but don’t expect to replace your 5G link anytime soon.

AWS ExtendDB 1.0: DynamoDB everywhere

Source: aws.amazon.com ↗

AWS announced version 1.0 of ExtendDB, an open-source, DynamoDB-compatible adapter for production use. It supports four storage backends - PostgreSQL, SQLite, Cassandra, and MongoDB - and runs DynamoDB workloads on-premises, in private/public clouds, or at the edge without code rewrites. For local dev and CI, the extenddb-dev container image and @extenddb/dev npm package provide a DynamoDB-compatible endpoint over SQLite.

Apache 2.0 licensed, maintained by AWS, developed openly on GitHub. This is the kind of thing that makes DynamoDB a portable API rather than a lock-in. If you’ve ever had to rewrite a codebase because you wanted to leave DynamoDB, this removes that cost. The SQLite backend for local testing is a small quality-of-life win that will save a lot of CI pain.

Foundation software: forks as the new PR

Source: jacobtomlinson.dev ↗

Jacob Tomlinson describes a shift toward “foundation software,” where developers fork existing tools and use AI agents to add personal features and maintain the fork against upstream releases. Agents can rebase changes and cherry-pick fixes, making forks sustainable for the first time.

The open-source implications are messy. 2026 has been a weird year: more code than ever, but also anti-AI stances and reduced community interest. The prediction here is that libraries remain critical, but simple projects get vibe-coded into personal forks. Ownership becomes personal rather than communal. Maintainers would need agents to search through forks for patterns worth merging upstream. It’s a plausible future, and it changes the economics of who maintains what.

Waymo's safety numbers don't add up

Source: azhdarchid.com ↗

A critique of Waymo’s safety record argues the claims are methodologically misleading. The comparison is to “all drivers” in counties where Waymo operates, despite Waymos doing little highway driving - which is statistically more dangerous. The author also notes Waymo’s peer-reviewed papers are all in the same journal, which raises questions about independent scrutiny.

The deeper argument is structural: tech companies can absorb fines and wield legal influence, so they have little long-term incentive to prioritize cyclist safety beyond a self-determined limit. For engineers building safety-critical systems, this is a reminder that the stats you present are as important as the ones you collect. If your comparison group doesn’t match your operating envelope, the numbers are marketing, not evidence.

Nvidia's interim CEO tries to steady the ship

Source: cnbc.com ↗

Colette Kress held her first all-hands since Jensen Huang’s resignation, telling employees to “let’s move forward”. The yearly review process moves from March to December. Board chair Harvey Jones said the board unanimously accepted the resignation and won’t reinstate Huang, despite employees reportedly wanting him back.

The stock is down 37% from its 52-week high. Moving reviews up a few months while the CEO search drags on is a signal that retention is the priority. When a company loses its founder and the stock drops a third, the first casualty is usually morale, and the second is talent. The December review cycle is a band-aid, not a strategy.

Gemini 4 Neon: cheaper and faster, but is it better?

Source: testingcatalog.com ↗

Google released Gemini 4 Neon on the Gemini API paid tier, Google AI Studio, Android Studio, Antigravity, and the Gemini app for Pro/Ultra subscribers. It costs $3 per million input tokens and $15 per million output tokens, with a 1-million-token context window. It scores 61 on the Artificial Analysis Intelligence Index and 65.4% on CursorBench 4.1, versus 58.1% for Claude Opus 5.5.

Neon has four thinking levels and claims 32% lower cost per task than Carbon at the default high level. For high-frequency coding workflows, that’s a meaningful price-performance shift. The CursorBench number matters: if it holds up in real workloads, this becomes the default model for agentic coding loops where token spend is the bottleneck.

The rest of the day

Source: aifoc.us ↗

Chrome extensions built entirely from prompts using the “prompt in a box” project, with a Studio UI and JSON schema for deterministic popup UI. Rapid prototyping, but don’t ship production extensions this way yet.

Agents speed up execution but not human synchronization. If sync takes 20% of your time, the max speedup is 5x even with instant execution. Smaller autonomous teams win.

Complex systems fail without “smart people who give a shit”. Recent failures at Microsoft 365, air traffic control, and major banks are attributed to a shortage of these people, not cyberattacks or AI. Capitalism drove them away, the author argues.

Scarcity has migrated from production to taste and judgment. A solo operator can ship in a weekend what took a funded team six months in 2019. The hard part is knowing what to build.

A Stripe employee banned slides and uses AI to make interactive presentations instead. Bauhaus-themed strategy explainers and planetary system org charts. Meetings are more fun. Your mileage may vary.

Get the brief

Liked this one? Pick the beats you care about and get just those, five minutes a morning. Free, no hype, one click to leave.

About Me Author

My name is

BriefTechNews

A daily digest of what actually moved in AI, tech, crypto and fintech, assembled and written with AI, with every item linked to its source. Read More
Tags

You May Also Like