AI Products

OpenAI

2026-10-07 02:38:41

OpenAI’s 28-day improvement pledge starts with a speed boost, but many users still want a reset

OpenAI product lead Thibault Sottiaux, known as Tibo, said the first delivery in his 28-day improvement pledge was a speed increase for GPT-6 Astra and GPT-6.1 Sol. According to his post, default speed for subscribers rose from 30 tokens per second to 50 tokens per second, and the change also applied to partner tools connected through Sign in with ChatGPT. He said users did not need to change any settings and should notice the improvement within about two hours. The update came after complaints around GPT-6.1 Sol, which launched on Sept. 29 and slowed during its first two days because of heavy demand. Tibo had said the model was close to being the highest-demand model in OpenAI’s history, apologized for the “slow start,” and said all paid accounts would be reset if the team failed to ship meaningful improvements each day over the next 28 days. On Day 2, Tibo announced four more updates, including free Auto-review with no quota charge, a simplified API paid-usage tier system, a meeting notes plugin, and the public beta of Decisions API. Even so, a poll he posted drew 74,565 votes, with 76% selecting “needs a reset” and 24% choosing “good day.” The report said no reset had appeared as of 10 a.m. local time mentioned in the article.

240

a16z

2026-10-06 14:01:37

a16z removes adult-focused AI products from its Consumer AI Top 50 rankings

Andreessen Horowitz, or a16z, has released the seventh edition of its Consumer AI rankings and changed the methodology in the process. The list, published every six months, tracks the most popular AI products across two segments: web products ranked by monthly visits and mobile products ranked by monthly active users, with separate Top 50 lists for each. In the latest edition, a16z said AI products primarily used for adult content will no longer appear in the rankings. The firm said the change was not driven by weak usage. It said that, under the previous methodology, those products would have taken more than 20% of the web Top 50, or at least 10 spots. Because of that, a16z decided to treat them as a separate category rather than rank them alongside other consumer AI products. The report also noted that mainstream AI assistants such as ChatGPT and Claude impose stricter limits on adult content, leaving room for demand for those products.

340

OpenAI

2026-10-06 12:05:30

OpenAI says GPT-6 Astra and GPT-6.1 Sol are about 50% faster on day one of its 28-day sprint

OpenAI has released the first update in its 28-day product sprint, with Product and Platform lead Thibault Sottiaux saying the default generation speed for GPT-6 Astra and GPT-6.1 Sol has been optimized by about 50% overall. The change applies not only to Codex and ChatGPT Work, but also to third-party products that use ChatGPT subscription credits through "Sign in with ChatGPT," including OpenCode, Pi, Amp, and Devin. Users do not need to change any settings, as the update will be pushed automatically. OpenAI had previously said that over the next 28 days it would either ship one update per day that brings a clear improvement for most Codex or Work users, or reset user credits directly. This speed upgrade marks Day 1 of that commitment.

250

OpenAI

2026-10-05 06:10:14

Tibo says OpenAI will spend 28 days on releases or resets as post-DevDay trust concerns grow

OpenAI executive Tibo said the company has now narrowed its near-term focus to four areas: simplifying products, improving efficiency to support more usage, breakthrough features, and new models. He later added that OpenAI is preparing for 28 consecutive days of releases or resets, with an emphasis on features that deliver obvious and practical improvements for users. The message did not land cleanly. After OpenAI’s recent DevDay, criticism from heavy users has become harder to ignore. While the event brought a long list of announcements, including GPT-6.1 Sol, new Codex / Work capabilities, Dots, and updates tied to agents, developer tools, and ChatGPT, some users said the releases did little to fix day-to-day issues. Complaints about lower usage limits, slower models, and a gap between pre-event hype and actual utility have fed a broader trust problem. The pressure is not only coming from users. Lauren Tan of the SpaceXAI Grok Bot team, known on X as @poteto, posted a pointed message aimed at OpenAI, prompting a quick response from Tibo. The exchange added to a wider debate over whether OpenAI’s product messaging is now running ahead of product quality.

350

OpenAI

2026-10-05 03:01:22

OpenAI sets a 28-day sprint for Codex, with daily visible improvements or a reset

OpenAI has put Codex and ChatGPT Work on a 28-day delivery schedule, according to Thibault Sottiaux, the company’s head of product and platform. He said the two products must either ship one update each day that brings a clear improvement for most users or go through a full reset of usage limits. A day earlier, Sottiaux said the team had already "locked in the target." He also said user feedback had become clear: people want the products to be simpler. From here, the team will focus on only four areas — simplifying the product, improving efficiency so users get more usable capacity, breakthrough features, and new models. The comments come as Codex faces stronger outside pressure. After Anthropic released Opus 5.5, community comparisons between Claude Code and Codex became more frequent, with most of the discussion centered on model capability, speed, and subscription limits.

340

OpenAI

2026-10-02 18:25:16

Sam Altman Says Dots Is His Favorite OpenAI Product So Far

OpenAI CEO Sam Altman said in a post on X that Dots is his favorite OpenAI product to date. He wrote that the product feels noticeably better every day as it learns more about his workflow and personal style. Altman added that Dots handles tasks he does not like doing, and said that makes him very happy. The remarks were attributed to Altman’s X account, @sama. The post did not include additional product details, launch information, or usage data in the source material provided. Techub carried the item as a brief update under its technology category.

390

ByteDance

2026-09-29 07:03:23

ByteDance’s Doubao moves ahead with personal assistant product plan as Spell nears release

ByteDance’s Doubao is pushing forward with product planning in the personal assistant category, according to a report carried by Phoenix Technology and cited from Sina Tech. People familiar with the matter said the company has continued the personal agent strategy it started with the Doubao mobile assistant last year, and began internal testing of an exploratory product codenamed Spell in April this year. The project was initially led by the Doubao mobile assistant team, but its external debut was delayed because of the Doubao phone launch and related factors. The report said Doubao is now accelerating the rollout of its personal agent capabilities as products such as Muse and Instinct draw attention overseas. Spell is expected to be tied more closely to the broader Doubao product, with the Spell project team and the Doubao dialogue team jointly responsible for the new product. A separate report by Dachang Ribao said Doubao’s product line received a new task during the Mid-Autumn Festival period to quickly rebuild a Muse-like product and deliver a demo version for internal review and testing by Sept. 30. The report added that teams across the product line and multiple upstream and downstream groups took part in the holiday development push.

770

Manus

2026-09-29 01:00:11

Manus founder says domestic version is in active preparation as product stays focused on a general AI agent

Manus founder Xiao Hong said the product has been positioned as a "general AI agent" from the start, rather than a tool limited to office work. In a public post, he said Manus was created before products such as Codex and Claude Cowork appeared on the market, and compared Manus to "the business of selling computers." In his view, users may buy a computer mainly for work, but the experience becomes narrow if that machine cannot also handle video or games, so Manus should not be confined to productivity tasks alone. Xiao also outlined how the team sees creation tools and infrastructure. He said video and games are important settings for unleashing creativity, and that these capabilities need to be improved through the local client Manus Studio. According to him, videos related to Cue were produced entirely with Manus Studio, without using a video generation model. Instead, the agent first wrote code and then turned that code into video. He added that, with cloud computer capabilities, Manus can also make online games that users can play together with friends. On infrastructure, Xiao said Manus cloud computers provide Linux, Mac, and Windows environments in the cloud. He also said the company is actively preparing a domestic version of Manus, while describing Cue as an entity with its own phone number, email, payment ability, computer, and sufficient intelligence.

570