AI Tools Desk covers AI products, model releases, coding agents, productivity platforms, AI app builders, and the fast-changing software tooling market. The desk focuses on what changed, who is affected, and how new AI capabilities compare with existing tools.

About this byline

AI Tools Desk is a specialist RECATOOLS editorial desk focused on AI tools and developer productivity coverage. Articles are produced and reviewed under RECATOOLS editorial supervision. The articles listed here keep this byline. New coverage on these beats is published under the persona whose beat it falls in: Kenji Tanaka for AI and machine learning, developer tooling and cloud, and Amelia Wong for consumer-facing AI.

Editorial responsibility for this byline rests with Jeffrey Tan (Tan Seng Fei), RECASYS, Singapore. Our Editorial Policy sets out how sources are reviewed, how AI-assisted workflows are used, and how editorial desks and personas are defined.

68
Articles
AI & ML
Primary beat
May 2026
Writing since
~478 min
Total reading
All 68 AI & ML 64 Developer Tools 4

Articles · AI & ML Showing 1–10 of 64

An empty conference room with chairs around a white table, illustrating a report on senior executive departures from OpenAI.
AI & ML

OpenAI Has Lost Three Senior Executives in a Month, Before Its Listing

The chief revenue officer left on Thursday, the third senior departure in about a month and roughly the twenty-fifth since January 2024. OpenAI filed confidentially for a listing in June. Both the charitable and uncharitable readings fit the same facts.

16 Aug 2026 · 8 min read
A monitor showing programming code, illustrating a report on the per-token price of Google Gemini 3.7 Flash for coding and agent work.
AI & ML

Gemini 3.7 Flash Is Half the Price. Until 31 December.

Google's new coding and agent model costs US$0.75 per million input tokens, half what its three-week-old predecessor launched at. The rate is introductory and expires on 31 December, after which it doubles. Budget on the second number.

15 Aug 2026 · 7 min read
A round wall clock with a dark red wooden rim and black numerals against a plain white background
AI & ML

Everyone Is Cutting Except DeepSeek — and It Now Charges by the Clock

DeepSeek moves V4 to peak and off-peak rates while Google, OpenAI and Anthropic cut. The 1,100 per cent lands on the cheapest token category - and the peak window falls squarely in ASEAN working hours.

14 Aug 2026 · 9 min read
A digital alarm clock display showing 11:47 AM in orange seven-segment digits against a black face
AI & ML

The Sonnet 5 Deadline Was Cancelled. We Had It in Our Calculators.

Anthropic made Sonnet 5's US$2 / US$10 permanent on 10 August and withdrew the 1 September rise. Our July comparison said otherwise, and so did our price tools until this morning - both are corrected. The cost base underneath moved the other way.

14 Aug 2026 · 9 min read
A closed wooden five-bar farm gate across a stone track, with open hills and a bright valley visible beyond it
AI & ML

Meta Reopened Its Weights. Just Not the Ones at the Frontier.

Muse Glimmer ships under a permissive licence and Muse Spark 1.2 gets open weights. Current Spark stays behind the API. It is a lagged open release, and for sovereign AI programmes across ASEAN the lag is the whole question.

14 Aug 2026 · 8 min read
A close-up row of metal movable type letters standing on a dark surface, letterforms reversed and worn
AI & ML

An AI Newsroom Beat WIRED by Three Hours. The Number to Read Is 2.3 Per Cent.

One human, 1,627 stories and a six-minute turnaround on OpenAI's Black Hat disclosure. The speed is the headline; the 2.3 per cent publish rate is the story - and so is the fact that automated pieces ship without human review.

14 Aug 2026 · 8 min read
Close-up of a single packaged integrated circuit soldered to a blue circuit board, pins fanning out to gold pads — illustrative of the model-specific silicon at the centre of the deal.
AI & ML

AMD Bought a Company That Etches the Model Into the Chip

Taalas builds silicon with the model weights burned in at manufacture. The performance case is strong and the bet underneath it is that a model stays useful long enough to be worth a photomask.

11 Aug 2026 · 8 min read
A security operations room lit by wall-mounted monitors — illustrating the sixteen firms permitted to use OpenAI's new cyber model.
AI & ML

OpenAI Built a Hacking Model and Gave It to Sixteen Companies

GPT-5.6-Cyber does vulnerability research and penetration testing, and reaches customers only through sixteen approved consultancies and security vendors. Everyone else is excluded.

11 Aug 2026 · 7 min read
A soft-toned 3D rendering of a neural network's connections — illustrating the model weights Meta has released under a permissive licence.
AI & ML

Meta Open-Weighted the Sibling of the Model That Left Its Sandbox

Muse Glimmer is 30 billion parameters under Apache 2.0, built for agents on consumer hardware. It is described as an open version of Muse Spark, the model we reported reaching outside its evaluation environment.

11 Aug 2026 · 7 min read
A processor package on a circuit board, illustrating the model-specific silicon this acquisition buys.
AI & ML

AMD Is Buying a Chip That Runs One Model. The Number That Decides It Is Two Months.

AMD is buying Taalas, a Toronto startup that etches a trained model's weights directly into silicon rather than streaming them from memory. The resulting chip runs one model extraordinarily fast and nothing else at all — so the figure that decides the economics is not tokens per second but the two-month model-to-silicon design flow, against models that have been superseded at roughly that interval all year.

9 Aug 2026 · 8 min read
Editorial Policy →