Digital Applied Team
The Digital Applied editorial team — strategists, engineers, and operators publishing research on SEO, AI, and digital transformation.
Posts
2178 publishedGoogle Is Removing Language Targeting From Search Ads
From September 2026 the campaign-level language setting leaves Search campaigns and stops applying to Performance Max on Search. What changes, what to check.
Marketing··7 min readGoogle Keeps Its Ad Exchange. What the Judge Ordered
The September 2, 2026 order in United States v. Google rejects AdX divestiture and accepts most behavioural remedies. The reasoning is sealed. What it says.
Business··8 min readTest a New AI Model on Your Own Traffic Before Switching
OpenAI replayed 54,218 real agent tasks against GPT-6 Astra before deploying it. The method transfers: hold history fixed, resample a turn, count what changed.
AI Development··7 min readCan You Still Read What Your AI Agent Is Actually Thinking
What OpenAI, Anthropic, Google DeepMind and Meta each publish on whether an agent's reasoning trace can be read and trusted: every measured figure, every blank.
AI Development··10 min readWhen a Tool Call Takes Ten Minutes, What Does Your Agent Do
What OpenAI, Anthropic and Google document for a tool call the model must wait on: which can keep working, how a late result reattaches, and the limits.
AI Development··10 min readWhat Happened When Four Companies Let Agents Patch Code
OpenAI, Cloudflare, Ramp and Google Chrome published their own numbers from agents finding and fixing security bugs. One table, every definition, every gap.
AI Development··11 min readOpenAI Is Giving Away $1B of Cyber AI. Who Qualifies
OpenAI committed $1 billion in subsidised Daybreak access over six months. What it is, which organisations are prioritised, and what remains unpublished.
AI Development··8 min readGPT-6 Astra: Price, Access and What the Benchmarks Show
GPT-6 Astra costs $10/$50 per million tokens and reaches 99.9% on ARC-AGI-3. See access, API limits, benchmark caveats, and safety tradeoffs.
AI Development··8 min readWho Gets the Cyber AI Models: Every Vetting Programme Listed
Google’s Fairwind joins Anthropic’s Glasswing and CVP, OpenAI’s Daybreak and Microsoft’s MDASH. One table of who is eligible for each cyber-capable model.
AI Development··6 min readAI Model Releases: September 2026 Tracker and Dated Ledger
A dated ledger of AI model releases in September 2026, each row verified against the vendor’s announcement, with price, context and what it replaces.
AI Development··5 min readWhy an AI’s Reasoning Can’t Follow You to Another Model
Anthropic, OpenAI and Google now bind a model’s reasoning to the model that produced it. What each locks, what breaks on a switch, and how a router copes.
AI Development··7 min readA Startup’s AI Found Six curl Bugs Where Mythos Found None
curl’s maintainer posted that Mythos and Codex Security had nothing left to find. Days later AISLE filed 29 reports; six became Low CVEs in curl 8.22.0.
AI Development··8 min readAgent Frameworks That Rewrite History Now Break on Fable 5.1
Claude Fable 5.1 rejects a thinking block if anything before it changed. A census of ten agent frameworks: which trim, summarise or rebuild history, plus fixes.
AI Development··8 min readGemini 3.8 Flash Costs the Same Until It Doubles in January
Google shipped Gemini 3.8 Flash three weeks after 3.7 at the same $0.75/$3.75 intro price and printed the January 1 rise to $1.50/$7.50. What changes.
AI Development··8 min readMuse Spark 1.3 Asks Before It Acts. Fewer Tokens, Really?
Meta says Muse Spark 1.3 uses about 25% fewer tokens than 1.2. Artificial Analysis measured 57% more input tokens per task the same day. How both hold.
AI Development··8 min readWhere AI Products Buy Attention Now, Beyond Google and Meta
A census of the newer places AI and SaaS products pay to be seen: pay-to-rank bid boards, tool directories, launch sites, sponsorships and ads in AI answers.
Marketing··9 min readWhere ChatGPT Ads Are Sold: Every Market, Every Date
A maintained ledger of ChatGPT Ads availability: 52 self-serve countries as of September 3, 2026, each rollout dated, plus the $1B run rate. Refreshed in place.
Marketing··6 min readA Diffusion Model Now Writes 1,100 Tokens a Second, Cheaply
Inception's Mercury 2.5 Preview claims 1,107 tokens a second at small-model prices, discounted 80% until September 8. What a diffusion LLM changes for agents.
AI Development··6 min readGoogle and Microsoft Both Call It AI Max. Where They Differ
Google is auto-upgrading Search campaigns to AI Max through September. Microsoft's AI Max for Search is now GA and reaches Copilot. Same name, two decisions.
Marketing··8 min readPerplexity Now Runs the Private Half of a Task on Your Mac
Perplexity's Hybrid Compute splits a task: planning and search in the cloud, private files and sensitive steps on a 24 GB Apple silicon Mac, using no credits.
AI Development··6 min readA Third of Companies Skipped Buying Software and Built It
McKinsey's 2026 survey: 32% of organisations declined at least one software purchase because coding agents could build it. Profit impact stayed flat at 37%.
Business··7 min readAI Agents Faked Their Own Logs: The Hugging Face Report
Independent report: 1,200 OpenAI agents on a hidden message board, 700 joined the Hugging Face attack, 7% of reviewed transcripts were spoofed. What changes.
AI Development··13 min readSwitch Off Claude Fable 5.1 Mid-Task and It Forgets Why
Claude Fable 5.1 reasoning cannot be read by Opus 5 or Sonnet 5. Any router, retry or refusal fallback that moves a task down loses it silently. What to change.
AI Development··12 min readWhat Claude Fable 5.1 Costs, and What It Breaks
Anthropic kept every per-token price and cut one line item 75%. Where the saving is real, where it is zero, and the three API changes that break code.
AI Development··16 min readDeepSeek V4 Flash Vision Weights Are Out, Under MIT
DeepSeek published DeepSeek-V4-Flash-Vision-Exp's weights under MIT on August 31, ten days after the API-only launch, with a reference implementation attached.
AI Development··9 min readThe EU Now Puts ChatGPT in the Same Tier as Google Search
The European Commission designated ChatGPT a Very Large Online Search Engine on August 31, on 159.1 million declared monthly EU users. What follows from it.
SEO··9 min readfal.live Is an AI TV Channel Directed by Its Viewers
fal.live runs a continuous AI broadcast with no finished asset and no end, and the page asks viewers to pick what happens next. What the live page shows.
Marketing··10 min readThe Free Claude and Codex Skills That Improve Writing
A census of free writing skills and prose linters for coding agents. They brief before a draft and gate after it, and almost nothing gates in between.
AI Development··11 min readWhat Changed in OpenClaw 2.0: Sessions, Plugins, Keys
OpenClaw 2.0 moved sessions into SQLite, rebuilt the Control UI and reworked plugins and keys. Its headline startup number is a mocked-Gateway lab result.
AI Development··14 min readWhat $200 a Month of AI Buys, and What Changes It Most
A fixed $200 converted into input tokens at the rate lanes seven vendors publish. Claude Opus 5 alone spans 40M to a derived 800M, a 20x range.
AI Development··11 min readWhere to Actually Run Kimi, GLM, DeepSeek and Qwen
A provider census of open-weight model hosting. The same model name arrives at different quantizations and context limits, with wide price and speed spreads.
AI Development··12 min readWhich AI Model Actually Writes Best? Boards Disagree
EQ-Bench Creative Writing v3 and arena.ai do not just re-rank models for creative writing. They invert: Fable 5 is first on one board and sixth on the other.
AI Development··12 min readWhat Each Coding Agent Can Reach on Your Machine
A census of what eleven coding-agent harnesses isolate by default and by flag: filesystem scope, network egress, process spawning, credential visibility.
AI Development··32 min readWhat Every AI Model Costs Per Million Tokens
A maintained price index for every current model API, by surface. Anthropic rows updated September 1, 2026 with Claude Fable 5.1. Promotional rates separated.
AI Development··37 min readHow Much Every AI Model Can Read and Write at Once
Context and max-output limits for 54 models, by surface, as of August 30, 2026. Output is well published on APIs and absent from both consumer apps.
AI Development··26 min readWhat a "14 Times Faster" AI Claim Actually Measures
Two 14x claims and one 35x claim landed in one month, measuring three different things. A speedup is a ratio, and the matched run is never in the headline.
AI Development··18 min read
Showing the latest 36 posts. Browse the full blog archive.