Cognition
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 14
- Study: People still deny AI true consciousness even when behavior matches humans
A team from LMU Munich's philosophy of mind and social neuroscience group published a study in the journal Cognition with nearly 1,100 participants in a controlled experiment: two groups read identical scenarios that differed only in whether the protagonist was a human or an AI, then rated the protagonist's awareness and consciousness. Awareness ratings rose with responsiveness regardless of whether the subject was biological or machine, but on consciousness the most responsive AI still scored below the least responsive human. Lead author Ophelia Deroy said people use psychological words like "confused" or "thoughtful" for AI only because models are trained to resemble humans and no better vocabulary exists, not because the public truly believes machines have minds.
科技新报 · 🔥 0 - Devin can now build and test iPhone Duo apps in macOS cloud VMs
Cognition said on X that Devin can now build and test iPhone Duo apps inside macOS cloud VMs, letting the agent unfold the phone, fold it, close it, check every screen and return video verification of the run. The examples given include Devin shooting someone's portrait while they watch themselves on the outer screen, sending posing cues, half-folding the phone so it stands on a table for timer shots, and closing it mid-countdown to confirm no photo is lost. The claims come from a demo post shared by Cognition's own account, with no independent verification and no details on availability or pricing.
cognition (X) · 🔥 0 - Cognition and National Design Studio launch American Forge government residency
Cognition said it has partnered with National Design Studio to launch American Forge, a six-month residency that places engineers inside the U.S. government to build software. According to the post, the program is run by National Design Studio in partnership with Cognition and is aimed at "exceptional engineers ready to take on the toughest problems in American government," with the stated goal of building software that affects hundreds of millions of people. Cognition said applications are now open and asserted that a small team can now accomplish more than ever, including in public service.
cognition (X) · 🔥 0 - Cognition to tie Devin usage limits to engineers' AI efficiency
To rein in AI spending, large companies are capping employee usage, switching to cheaper smaller models and using routers to send only the hardest requests to the most expensive models. Cognition told The Information exclusively that it is adding a feature that analyzes an engineer's coding session in Devin, estimates the equivalent human hours the work would have taken, and divides by the session's cost to produce an efficiency measure — so companies can steer more resources to their most efficient employees. Calvin Giroud, a Cognition vice president of customer engineering, said user limits should not be static but dynamic based on usage efficiency and the value obtained.
The Information · 🔥 0 - Modal raises $750M Series D at $15.75B valuation, tops $1B ARR
AI infrastructure company Modal said it raised a $750 million Series D at a $15.75 billion post-money valuation, led by Accel with Kleiner Perkins joining and existing investors including Menlo Ventures, General Catalyst, Redpoint, Lux Capital and Bain Capital Ventures participating. The company said it has crossed $1 billion in annualized run rate, that usage has tripled in the five months since its Series C and grown more than 15x year over year, and that its Sandboxes product alone runs at roughly a $300 million annual rate with close to 3 billion sandboxes started since 2023. It lists customers such as DoorDash, Ramp, Linear, Cognition and Runway, alongside robotics firms Physical Intelligence and 1X training robots on its platform.
Modal Blog · 🔥 0 - Cognition's Devin integrates with Pylon to auto-fix tickets
Cognition announced a partnership with customer support platform Pylon to bring AI coding into the ticketing workflow: Devin can automatically fix bugs and build features as customers report them. Cognition says Pylon tickets now get fixed automatically by Devin, that new Devin customers on Pylon receive $15,000 in credits, and that a webinar will show the integration in action. These are promotional claims from Cognition and Pylon; independent verification of scope and results is not provided in the sources.
cognition (X) · 🔥 0 - Ed Zitron: AI bubble built on inflated annualized revenue figures
In his Where's Your Ed At newsletter, Ed Zitron argues the AI boom is propped up by "stupid, made up numbers" — annualized revenue figures that make AI startups look more successful than they are — and calls their popularization a failure of journalism. He cites Financial Times reporting that OpenAI's annualized recurring revenue, reported by Axios at $70 billion and confirmed by the FT, was actually about $50 billion, with the higher number "grossed up" to include sales through cloud partners, while Anthropic books gross revenue on cloud-partner sales and told investors in August it had reached a $65 billion annualized run rate based on the prior 28 days of consumption revenue plus subscriptions. Zitron also contends AI startups largely resell OpenAI and Anthropic tokens at steep discounts while running at a loss funded by venture capital, so each new model release produces a revenue bump that could fade if that funding tightens.
Where’s Your Ed At · 🔥 0 - Devin adds Dynamic workflows to split jobs across multiple agents
Devin Developers said in an Oct. 13 post on X that Dynamic workflows let Devin split a large job into pieces, hand each piece to its own Devin instance, and bring the results back together. The account said the plan for a Dynamic Workflow lives in code rather than in an agent's head, so users know which Devins will run, in what order, and what each one is doing. The post did not state an availability date, rollout scope or usage details.
cognition (X) · 🔥 0 - Personal AI assistant startup Tab exits stealth at $300M valuation
TechCrunch reported that US personal AI assistant startup Tab emerged from stealth on October 6 at a $300 million valuation, with SV Angel, Valar Ventures and American Spirit among its investors; the company did not disclose the actual amount raised, nor its user numbers, revenue or paid usage. Tab is operated by Terrasoft, a Y Combinator Winter 2020 batch company, and users can send instructions via iMessage or WhatsApp to order food, plan trips, buy gifts, make calls or pay bills; it has its own phone number, computer and wallet, and can connect accounts including Gmail, Google Calendar, Slack, GitHub, Linear and Notion, though it remains in closed beta. In a 15-task test by the independent Assistant Benchmark site, Tab received 97 messages over two days with a median reply time of 12 seconds, scoring 5.9 overall and ranking 21st: it earned 10 for a phone task and 9 for permissions and privacy, but only 3 for restaurant booking by web and phone, 5 for a travel task and 1 for a purchase task, where card payments failed repeatedly. The leaderboard's top three were Meta Muse at 9.3, Instinct at 8.5 and OpenAI Dots at 8.4.
36氪 人工智能 · 🔥 0 - Devin can spawn managed Devins, forming a multi-level agent tree
Cognition's Devin Developers account said Devins can start managed Devins of their own, so one task can branch into a tree of Devins several levels deep. According to the post, work split this way takes roughly as long as its slowest part rather than the sum of all parts, meaning task size stops being a reason to postpone work. No version, rollout or availability details were provided.
cognition (X) · 🔥 0 - SGLang and Miles tuned for NVIDIA Vera Rubin, serving Kimi K3 NVFP4
The SGLang & Miles team says it received early access to two NVIDIA Vera Rubin nodes (8 GPUs) and brought both SGLang inference and Miles RL training up on the platform. For serving the Kimi K3 NVFP4 checkpoint, it retuned attention, MoE/communication and KDA kernels, reporting a 16% faster deeper MLA pipeline at batch 16 and 128k context, a 20% faster split-KV reduction, a 5.9% end-to-end gain from fusing the MoE tail, and 276 fewer kernel launches per decode step. On the training side, Miles uses SGLang for rollout and Megatron for training, and ran end-to-end RL out of the box on a single 4-GPU tray with Qwen3-30B-A3B and DeepSeek-V4-Flash, plus agentic RL with 64 sandboxes on the Arm-based Vera CPU on SWE-bench Verified. The team also cites Cognition as reporting up to a 4.8x increase in total token throughput on Vera Rubin NVL72 versus GB200 NVL72 using its own private SGLang fork.
LMSYS Blog · 🔥 0 - AI's simmering question: where is all the automation?
A The Information column frames the industry debate over why automation has not arrived even as OpenAI said its unreleased model solved or made major progress on hundreds of significant math and computer science problems. Cognition founder Scott Wu argued that ever-more-capable agents plus more data centers, memory and sandboxes will yield "virtual employees," while ex-OpenAI researcher Diogo Almeida, now at TypeSafe AI, maker of Jev, said today's LLMs are optimized for assistance and fail at unattended automation. Anthropic's Claude Code product lead Cat Wu conceded much automation stalls at a semi-automated stage where Claude does 80% and a human does 20%.
The Information · 🔥 0 - Devin gives every session its own cloud Mac for iOS work
Cognition's Devin Developers account said cloud agents become far more useful when they can run the real toolchain: xcodebuild, the iOS Simulator, and Computer Use tapping through the UI. It said every Devin session now gets its own cloud Mac, so mobile work can be delegated like any other ticket.
cognition (X) · 🔥 0 - SemiAnalysis: Anthropic subscriptions offer ~5x OpenAI's API-equivalent value
SemiAnalysis estimates that at the mid-tier model level — Claude Opus 5.5 versus GPT-6.1 Sol — Anthropic subscriptions deliver roughly 5x the API-equivalent value of OpenAI's, while top-tier limits are broadly similar across GPT-6 Astra and Fable 5.1. It estimates Anthropic's subscriptions are only ~10% of revenue but consume over 40% of inference compute, cutting blended revenue per MW by about $36M. OpenAI last week halved the API-equivalent value of its $200 tier and added a $500 tier that offers only 21% more Astra usage than the old $200 plan, with pre-cut $200 plans keeping old limits until October 29; SemiAnalysis also launched a subscriptions dashboard tracking OpenAI, Anthropic, Meta, SpaceXAI, Cursor, Cognition, Z.ai, MiniMax and Moonshot.
SemiAnalysis · 🔥 0
Experience and discussion from the community
Share my Cognition experienceAsk about Cognition
Nobody has shared their experience with Cognition yet.