AI Products news and analysis on Web Pulse: 23311 curated articles tracking the latest AI Products developments, tools, and research, updated continuously from vetted sources.
A hyper-optimized, zero-dependency C/CUDA inference engine for the Qwen 3.6 35B model on RTX 5090 Blackwell GPUs achieves 13.4k tokens/sec prefill throughput at 2,048 context depth and 270+ tokens/sec decode, outperformi…
A developer has built Mascot Engine, a system that connects interactive Rive mascots to real product states, enabling them to react to user actions, AI responses, errors, and more at runtime. The mascot's behavior is con…
Apple is reportedly preparing an aggressive countermeasure to OpenAI's competitive threat, with internal plans described as 'thermonuclear' in scope. The move signals escalating tensions in the AI industry as Apple seeks…
A developer benchmarked GPT-5.6 against Claude Code and OpenCode, finding GPT-5.6 delivers a 2.2x speed advantage over OpenCode and reduces costs by 21% compared to Claude Code. The tests measured speed, latency, cost, a…
A 50-year-old open conjecture in graph theory, the Cycle Double Cover Conjecture, was proved by 64 subagents coordinated in roughly one hour on 2026-07-12, demonstrating a routing layer where one model calls many models.…
Cockroach Labs launched three new hands-on courses on Cockroach University that teach developers and database administrators how to use AI agents to automate CockroachDB operations such as rolling restarts, encryption ke…
AI models retain between high single digits and 40% of customers after five months, with the stickiest foundational cohorts near the top of that range, according to a study by OpenRouter and a16z. The average frontier mo…
OpenAI released GPT-5.6 Luna, Terra, and Sol, which cleared the 'High' cybersecurity capability threshold under OpenAI's Preparedness Framework, indicating the models can automate end-to-end attacks and vulnerability dis…
Using GPT-5.6 Sol as an orchestrator with cheaper sub-agent models like Luna or Terra can reduce per-run costs by 60-80% while maintaining output quality, according to a guide on multi-agent orchestration. The approach s…
A 2026 pricing comparison of GPT-5.6 Sol, Grok 4.5, Meta Muse Spark 1.1, and Claude Fable 5 shows that GPT-5.6 Sol costs $1.80 per million input tokens and $7.20 per million output tokens in standard mode, with extended …
ChatGPT Work Mode, available on Enterprise and Teams plans, can be used to build a personal productivity dashboard that monitors email, Slack, calendar, and brand mentions, compressing morning orientation time from 45-90…
Claude Code users running the Fable 5 architecture now pay API rates instead of a flat subscription, with input tokens consuming 60-75% of total usage and system overhead accounting for 15-30% of input tokens. To cut cos…
Real founders built six- and seven-figure AI automation businesses without coding, with case studies showing revenue from $25K to $2.7M ARR. Marcus, a bookkeeper, grew a solo operation to $25K ARR in four months by offer…
MindStudio has joined Upwork's Partner Certification program, making its Certified AI Agent Builder credential recognized on the platform. Certified freelancers can display a verified Partner Certified badge on their Upw…
Conductor 0.75.0 adds GPT-5.6 Luna to the Codex model picker, introduces a PR badge with an Open in GitHub action, and improves agent review comment reading. The update also includes a new terminal engine for persistent …
Production finance agents at Artifi operate as a team of narrow specialists with explicit permissions and institutional knowledge, not as a single general agent, according to a blog post by the company. The bill-processi…
Vercel's AI Gateway leaderboards now offer open data under CC BY 4.0 and shareable chart images, ranking models, labs, apps, and providers by production traffic aggregated daily across trillions of tokens. Users can down…
A study by Maximilian Weber and Johannes B. Gruber analyzing Twitter data found that Meta's Galactica, marketed as a scientific assistant, faced rapidly escalating negative sentiment after its November 2022 release, whil…
Lemlist, an AI-powered cold outreach platform with over 37,000 business customers, sends hyper-personalized emails that bypass standard spam filters by passing SPF, DKIM, and DMARC authentication, warming sending domains…
FrontierFinance, a new benchmark from an unnamed source, evaluates AI research systems on 220 open-ended financial research queries using 11,543 expert-authored rubric criteria across six workflows. The top-performing sy…