Pondero AI logo

Pondero AI

Archives
Log in
Subscribe
October 1, 2026

Pondero Brief: Google's Gemini 4 Argon claims 77.9% on DeepSWE, pilot partners only

Pondero Brief - OCTOBER 1ST, 2026

Plus: CustomGPT.ai's "1,000 queries" are credits, and we cut our rating to 3.8.  ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌
pondero. BRIEF · OCT 1

Google's Gemini 4 Argon patches vulnerabilities, but you can't use it yet

1M-token output and a claimed 77.9% DeepSWE score, gated to Google's Fairwind pilot

Google shipped Gemini 4 Argon on September 30: 1M-token output (16x the prior generation) and a claimed 77.9% on DeepSWE v1.1, per its announcement. Only vetted Fairwind Program defenders can run it; intro pricing is set but there is no public release date.

Also in today's brief

  • CustomGPT.ai's 1,000-query plan is actually credits
 
Quick Hits
• DoorDash put an ordering agent inside Apple Messages. Text "order my usual," get local picks with photos, or split a group order across dietary needs, per TechCrunch. Waitlist only, US only, no pricing disclosed: for consumer-commerce teams, the chat thread is now a checkout surface to track.
• Meta says Muse cannot read your Messages by accident. A columnist said Meta's Muse Mac agent read his private messages without permission. Meta VP Andy Stone told TechCrunch the integration is "entirely opt-in," gated behind Full Disk Access plus a separate Messages connector, steps that "cannot be circumvented accidentally." Before a desktop agent goes on team laptops, audit who has granted Full Disk Access, because that macOS toggle is the permission boundary that counts.
From the Pondero Stack
CustomGPT.ai review illustration

CustomGPT.ai's Standard plan buys about 100 verified answers a month, not 1,000.

The pricing page says 1,000 queries, but the product bills credits, and Verify Responses costs 10 per answer on the $99 Standard plan versus 2 on Premium, per CustomGPT.ai's actions-cost doc. Our July review missed this, so the rating drops from 4.0 to 3.8: a CX team that needs verified, cited answers should budget for Premium at $499/mo, and Standard fits one quiet internal bot left at default settings. See the credit math tier by tier or try CustomGPT.ai.

 

How was today's brief?

★★★★★ Nailed it  |  ★★★ Solid  |  ★ Missed

Jonathan Hildebrandt Jonathan Hildebrandt
Co-founder and primary operator of Pondero. Writes the Pondero Brief.

Affiliate disclosure  ·  Unsubscribe  ·  Manage preferences

Pondero earns commissions on some links. This does not affect our editorial picks.

Don't miss what's next. Subscribe to Pondero AI:
← Newer Pondero Brief: Cline lands on Linux; pin your model before it shifts again Older → Pondero Brief: Sonnet 5.5 beats Opus at half price, misses more bugs
pondero.ai
Bluesky
LinkedIn
Twitter
LinkedIn
Powered by Buttondown, the easiest way to start and grow your newsletter.