DoorDash is moving to reshape its role in the food and retail ecosystem, unveiling on Tuesday a series of features that borrow directly from the worlds of social media, loyalty programs, and artificial intelligence.

The company announced a creator program to pay users for producing short-form videos of their meals, rolled out “Going Out” dining rewards, and introduced AI-powered personalization tools.

The creator program, open initially in 20 U.S. cities such as Atlanta, Austin, Miami, and San Francisco, will compensate participants who record videos of local dishes. The goal is to bring discovery into the app, letting users preview meals before they order. Uber Eats piloted a similar TikTok-style feed last year, underscoring how delivery apps are racing to embed content into their platforms as a way of keeping users engaged. DoorDash has not yet detailed its monetization structure for creators, but said the initiative will expand nationwide through the end of the year.

Another step is Going Out, a feature that rewards customers for dining in restaurants. DashPass members can earn exclusive benefits and redeem offers when eating out, with early testers receiving an average of $9 in value per order. Rewards are now live at thousands of restaurants across the U.S. and Australia, and for a limited period, even non-members will gain access.

“With Going Out, you can find and redeem exclusive in-app offers when you dine in and earn rewards just for coming back,” said Parisa Sadrzadeh, vice president of strategy and operations, at a press event on Monday. “This year, Going Out customers received an average of $9 in value per order when using offers.”

The feature positions DoorDash in direct competition with companies like OpenTable and Resy, especially as it integrates with SevenRooms, the hospitality platform it acquired earlier this year, to allow in-app reservations.

Artificial intelligence is also playing a bigger role. The DoorDash app now generates personalized recommendations based on user history, budget, dietary needs, location, and even time of day. At checkout, a “Complement your Cart” section proposes add-ons to simplify grocery shopping.

Meanwhile, new smart tags help users filter restaurant menus by identifying dishes as vegetarian, gluten-free, high-protein, or spicy, drawing from reviews, text, and photos. These tags are already live in the U.S., Canada, Australia, and New Zealand.

“We’re looking at an individual restaurant item, and we use every piece of information we have. The customer reviews, the text from the merchant, the photo of the item, to start to infer things about those items,” Austin Haugen, DoorDash’s VP of product, said at the event. “So we can learn this item is vegetarian, this one is gluten free, spicy, high in protein, etc. Once we infer this about the items, you’ll start to see these tags around the app.”

Beyond food, the company has expanded its comparison shopping tools to cover categories like beauty, electronics, and pet supplies, building on a feature previously limited to alcohol.

The strategy reflects an intensifying battle among delivery giants to capture not just orders but consumer attention and loyalty. Uber Eats has leaned on cross-platform perks with Uber’s rides and has been experimenting with its own video-driven discovery features. Grubhub, meanwhile, has turned to partnerships, such as its tie-up with Amazon Prime, to expand its equivalent of DashPass.

Each company is trying to redefine itself as more than a logistics network: Uber Eats as an integrated lifestyle platform, Grubhub as a membership-driven service, and DoorDash as a hybrid of marketplace, media hub, and AI-powered shopping assistant.

OpenAI generated about $4.3 billion in revenue in the first half of 2025, according to financial disclosures to shareholders reviewed by The Information. The figure represents a 16% jump over its total revenue for all of 2024, underscoring the scale of demand for artificial intelligence tools like ChatGPT.

At the same time, the disclosures show the AI company is still spending aggressively to secure its lead. OpenAI reported a cash burn of $2.5 billion in the first six months of the year, largely driven by the immense cost of running and developing large-scale AI models.

Heavy R&D Spending

The report highlights that OpenAI poured $6.7 billion into research and development during the period, a level of spending unmatched in the sector. Even after such outlays, the company closed the half-year with $17.5 billion in cash and securities, giving it substantial firepower to pursue expansion despite mounting costs.

OpenAI told investors it expects to meet its full-year revenue goal of $13 billion while keeping cash burn within its $8.5 billion target.

Valuation and Liquidity Moves

The financial disclosures land as OpenAI continues to explore ways to broaden its capital base. In August, Reuters reported the company was in early discussions about a stock sale allowing employees to cash out, a move that could value the firm at roughly $500 billion. Such secondary sales are increasingly common among late-stage private companies as they balance high valuations with the need to retain talent.

Investor appetite remains strong for Nvidia, which has benefited enormously from the AI boom. The company announced last week that it will invest up to $100 billion in OpenAI and supply data-center chips critical to scaling the company’s operations. The partnership links OpenAI more closely to the dominant supplier of AI infrastructure, ensuring a steady pipeline of computing power even as rivals like Anthropic, Google DeepMind, and xAI expand their own model development.

How OpenAI Stacks Up

While OpenAI’s revenue trajectory is the steepest in the industry, its burn-to-revenue ratio is far higher than its peers. By comparison, Google DeepMind — folded into Google’s broader AI division — is estimated to spend several billion annually, but Alphabet’s vast advertising revenues cross-subsidize those costs, shielding investors from the kind of direct burn rate OpenAI is shouldering.

Anthropic, backed by Amazon and Google, has pursued a more measured growth strategy. Its last reported figures suggested annualized revenue in the low billions, but with lower infrastructure costs as it leaned heavily on its cloud backers for credits and partnerships. xAI, Elon Musk’s entrant, has raised billions but is still at an earlier stage of monetization, focusing on consumer subscriptions and integrations with Tesla and X.

In this context, OpenAI’s $6.7 billion R&D spend in just six months signals both its scale and its structural disadvantage: unlike DeepMind, it lacks a parent company with diversified revenue streams, and unlike Anthropic, it must sustain massive compute costs more directly on its balance sheet.

The Assumptions

For financial markets, OpenAI’s numbers highlight the paradox of generative AI: surging top-line growth paired with eye-watering costs. Investors are expected to interpret the $4.3 billion revenue haul as validation of enterprise adoption and ChatGPT’s stickiness, but the $6.7 billion R&D spend underlines how capital-intensive the race for model dominance has become.

Behaviorally, the disclosures may harden a view among investors that only a handful of players — those with access to vast computing resources and strategic partners like Nvidia — can realistically compete at scale. At the same time, the looming $500 billion valuation could spark debate in financial circles about whether OpenAI is entering “mega-cap tech” territory in private markets, or whether its cash burn points to a model that will remain structurally expensive for years.

With OpenAI signaling confidence in its full-year revenue and cash-burn targets, markets may treat its disclosures as evidence that the AI wave remains in full force — but also as a reminder that profitability, not just scale, is the long-term hurdle.

Tekedia Capital joins Paul Graham (a founder of YCombinator), executives in OpenAI, and other global investors to seed Nozomio, with $6.2 million, under the solo and singular one-person coding prodigy Arlan Rakhmetzhanov.

Arlan’s story is a magical one: from Kazakhstan to Silicon Valley for an advanced context augmentation system for AI agents . We wish this secondary school dropout the best of luck as he takes Nozomio to the next level.

I started Nozomio in high school six months ago as an AI agent for codebases that could cross-reference multiple repos, answer questions, and help developers onboard themselves. I failed.

But I knew there was something in the agent space that every developer wants: the right context.

Right before my Y Combinator batch, I was struggling to give coding agents the right context. They missed versions, filled up the context window, and used incorrect frameworks. That is when I built the first iteration of Nia, which is now an augmentation layer for AI agents.

Context is the most important problem in the agent space. It prevents AI systems from achieving superhuman code intelligence, and I want to solve it.

The current state of Nia might change a thousand times, but the problem will stay the same. Even as an MCP, which is very novel and underdeveloped, Nia has already improved every coding agent by at least 30%. I am on the mission to 100x it.