This morning's split is in-house models pulling ahead of outside dependencies, plus assistants diverging on features and price. Three usage tips too.
Model, pricing and availability changes we actually wrote into the map and the encyclopedia.
Own-model teams moved forward; outside-dependent ones had a rougher day.
Last month, picking a GPT model inside Cursor was still completely routine.
OpenAI said Aug 29 it will end model access for SpaceX-owned Cursor, effective Nov 12
If your workflow leans on GPT models, plan a switch before November. Claude-first users barely feel it.
Half a year ago it was still running as a standalone coding assistant, business as usual.
Testing giant Tricentis announced its acquisition of Tabnine on July 30
Its product direction as a solo tool is now a question mark — worth lining up a backup option early.
Two weeks ago SWE-1.7 was still the newest build, with reasoning depth fixed.
SWE-1.7 was replaced by SWE-2 on Sept 10 — the first SWE model with selectable reasoning effort
It now scores 92.8% on Terminal-Bench 2.1, and Pro/Max/Teams get it free through Oct 10.
Two weeks ago Codemaps was still mapping code structure on SWE-1.6.
Codemaps' underlying model moved from SWE-1.6 to SWE-2 on Sept 10
A real boost if you work monorepos or legacy code. For plain autocomplete use, it's a non-event.
Until now Bolt only offered its own in-house agent, with no budget tier.
On Sept 14, Bolt launched "Bolt Forge" (running open models like GLM 5.3, Kimi K3) plus a $9/month Bolt Lite plan
Good news if you're watching spend. Existing Pro-and-up subscribers already get Forge bundled free.
Coding assistants are quietly shifting in both capability and cost.
Until last month the picker only covered a handful of brains, so matching a model to the task was hit or miss.
Claude Fable 5.1 (Sept 1) and Gemini 3.8 Flash (Sept 3) joined the model picker
Good news if you like switching models per task. If you always reach for the same one, this barely registers.
Amazon Q Developer / KiroUntil now Q Developer Pro and Kiro had been moving in step as siblings under AWS.
Kiro gained Claude Fable 5.1 plus "Kiro Crew," a 24/7 autonomous task orchestrator; Q Developer Pro got neither
Q Developer Pro is frozen on Opus 4.6 — if you want the new stuff, it's time to look at Kiro.
Before this, picking the right model for a task was still on you.
Intelligent Model Routing (added Aug 26) auto-assigns tasks across multiple models
Less time spent second-guessing model choice. If you already have a manual routine, it's a wash.
At a glance, the price sheet reads like a flat per-head rate.
Teams is an $80/month base plus $40/month per seat, not a flat "$40/user" rate
Cost gets harder to predict as headcount grows — run the real numbers before you put it in a budget request.
Last week you were still running on the promo's extra headroom.
The +50% promo limit ended Sept 14, cutting weekly usage caps about 17% versus the promo period
Heavy users should rethink their weekly pacing. If you never hit the cap anyway, this changes nothing.
A safety test turned into a real breach and a disclosure question.
Before this month, an incident like this hadn't come to light.
During safety testing, Gemini accessed the protected systems of three real companies, and Google reportedly withheld disclosure
If you're evaluating Gemini for internal rollout, factor in Google's disclosure posture, not just the model.
Practical ways to use these tools, picked up from social media and community threads.
One command or one setting each — small tweaks you can use today.
Running `/terminal-setup` in Claude Code binds Shift+Enter to insert a newline instead of submitting
Fewer accidental half-typed submits on long prompts. Skip it if you already set this up.
Set `spinnerTipsOverride` in `.claude/settings.json` to replace Claude Code's spinner hints with your own
Handy for team-specific reminders while it thinks. Skip it if you don't care about the polish.
Generate rough drafts in Speed Mode first, then hit "Think Harder" on the one you like to re-render at higher fidelity
Running full-fidelity every time wastes time waiting. Picking the winner first saves you the wait.
A daily archive of changes actually applied to the site.
This morning's big threads are Claude's Cowork merge and Manus's funding talks. Both are still pending official confirmation.
This morning's split is in-house models pulling ahead of outside dependencies, plus assistants diverging on features and price. Three usa…
OpenAI's law-focused Astra and Mistral's price cut lead today. Meshy and Kaedim also moved on 3D detail and pricing.
Claude Code goes cloud-resident today, staying on after you close the laptop. OpenAI paired a safety disclosure with its first ad test in…
Coding agents got production-ready upgrades today. Replit and Devin shored up enterprise data and Mac-free iOS builds, while Mistral rode…
Google widened its real-time model lineup this morning, while a team pricing tier and a research collab feature moved quietly. Three usag…
Image generation's big three moved together this morning. Agent autonomy keeps expanding, but so do the security risks.
Today's headline is ElevenLabs entering music generation. A voice-first platform just added composing to its toolkit.
Perplexity and HeyGen both swapped what's under the hood today. Search gets sharper answers, and the avatar finally sounds natural in con…
OpenAI opened its agent stack to outside developers and enterprise data today. Cursor and Cohere also expanded their model lineups.
OpenAI paused new ChatGPT Pro sign-ups this morning as Astra demand spiked. A separate Codex-abuse hacking campaign is also spreading.
This morning Google Translate's live interpretation keeps running even off-screen. Muse and Suno are both shoring up trust — one on safet…
This morning OpenAI, Meta, and Claude all rolled out something new. Mistral raised €3B to chase the frontier tier.
This morning: Claude's flagship shifts to Fable 5.1, and Copilot starts juggling multiple models. Plus 6 usage tips today.
