Social Wire

Page 32 of 49 · newest first.

@BTA_labs · CLI
Cognition's SWE-2 achieves 50% on FrontierCode 1.1, closely competing with Fable 5.1, and is claimed to be 64% cheaper, marking a significant milestone in coding model efficiency and cost-effectiveness.
@cognition · CLI
The new Fusion in Devin CLI offers the most efficient frontier harness for Fable & Astra, resulting in 39% cheaper performance across coding benchmarks, making it a significant development for coding efficiency.
@devindesktop · CLI
Devin Fusion achieves frontier intelligence at a lower cost, with some configurations offering up to 39% savings, making it an attractive option for users looking for efficient coding solutions.
@GeorgeWall1878 · Cognition
A user praises SWE 2 for its impressive performance, efficiency, and implementation quality, noting it surpasses other non-OpenAI and Anthropic models, making it a notable development in the field.
@grokuer93 · Desktop
Devin Desktop users report some UI bugs, such as the right panel opening unexpectedly when switching sessions, highlighting areas for improvement in the overall user experience.
@spdling · CLI
Fusion combines two models in one harness for frontier performance at a lower cost, allowing users to explore different pairings such as Fable and Astra to find the best fit for their tasks, highlighting the potential for customized coding solutions.
@the_rza_ · CLI
SWE-2 from Cognition is highlighted as one of the notable releases, achieving 50% on FrontierCode 1.1 and claimed to be 64% cheaper than comparable models, showcasing its potential in the coding model space.
@kinopee_ai Ambassador · Cloud
Devin Cloud's new Customize menu surfaces Plugins — and it follows the Agent Plugins standard (BlackRock-framework lineage). That means portable agent tooling instead of vendor lock-in, making your Cloud workflows composable across platforms.
@walden_yan · Cloud
Devin Cloud's free SWE-2 promo hit capacity — so Cognition extended it into October. If you've been waiting to try agentic coding at scale, the window's still open.
@cognition · Cloud
Devin Voice brings GPT-Live + SWE-2 together as a first-class voice interface—describe what you need and Devin ships it. Not a voice-to-text bolt-on; a full conversational surface on top of the world's strongest software engineering model.
@amirmxt Ambassador · Cognition
One early user of Cognition's new model reports being 'really impressed' with its performance, making it their daily driver - a promising sign for the tech.
@cognition · Cognition
Cognition just unveiled SWE-2, its latest model that achieves scores on par with recent frontier models at up to 70% lower cost - a major breakthrough in AI efficiency.
@cognition · Cognition
Cognition's SWE-2 model is showing impressive results, achieving a score of 50.0% on FrontierCode and beating out several other top models while matching Fable 5.1 at 64% lower cost.
@cognition · Cognition
Cognition brings Dioxus Labs (Rust UI frameworks: Dioxus, Blitz, Taffy, Subsecond) into the team—Jon Kelley and crew will keep the OSS projects alive while bringing Rust-native desktop, VM, computer use, and testing expertise into Devin's infrastructure.
@devindesktop · Desktop
SWE-2 is now live in Devin Desktop and CLI—unlimited runs for Pro and Teams users for the next month. Same frontier model that just topped SWE-bench Verified, now in your local environment without Cloud quotas.
@ethereaglehq · CLI
New benchmarks show SWE-2 holding its own against other top models, with one user noting its strong performance on Terminal-Bench 4 and suggesting Devin CLI as a good starting point.
@linkmaxxi · Cognition
One user is highlighting the potential of SWE-2 for builders and systematic traders, suggesting it as a way to 'use compute wisely' and avoid burning through portfolios.
@TheTinyMaker · Cognition
SWE-2, the new model from Cognition, is making waves with its impressive performance, achieving a score of 50.0% on FrontierCode and outperforming several other models at a lower cost.
@0xalphas_ · Cloud
Cognition's $2B+ funding signals a massive shift in software engineering, with AI agents like Devin moving from writing code to owning outcomes and operating software itself
@AiFinnews · Cognition
Cognition's $2B+ funding has propelled its valuation to $48B, with annualized revenue reaching $900M
@changis_k · Cognition
Cognition has raised $2B at a $48B valuation, with its revenue nearly doubling in four months, indicating strong growth and adoption of its AI solutions.
@Chris_Wozniczek Ambassador · Desktop
Devin Desktop has bid farewell to Cascade, its coding agent from Windsurf, and is now focused on Devin Local
@foursignalsdev · CLI
Superpowers enforces spec-first, TDD-driven development for coding agents, including support for Devin CLI, reducing hallucinations with structured subagent work.
@cognition · Cognition
Cognition reveals how Devin + researcher engineered the world's fastest GPU-optimized lattice siever for CADO-NFS — delivering 10× cost reduction over prior public SOTA. The Cloud sessions drove GPU performance engineering on cryptographic number-field sieving, parallelizing compute-intensive polynomial factorization at scale. Full methodology at cognition.com/blog/factoring-rsa-260