NoCode Weekly: June 28 — OpenAI GPT-5.6 Sol Preview, U.S. Loosens Mythos Restrictions, DeepMind Exodus Continues

TL;DR

OpenAI previews GPT-5.6 Sol — next-gen model with stronger coding, science, and cybersecurity capabilities. Rollout limited to trusted partners after a U.S. government request [1][2].
U.S. loosens restrictions on Anthropic’s Mythos 5 — Commerce Department allows limited return of Mythos to 100+ trusted U.S. organizations [4][5].
Google DeepMind talent exodus accelerates — Noam Shazeer heads to OpenAI; Jonas Adler and Alexander Pritzel leave for Anthropic. Alphabet shares slid on the news [7][8].
Zapier shifts AI steps to model-based pricing — Starting June 15, AI steps are priced by model tier (Standard, Advanced, Premium), with the default tier consuming 3x tasks per step [11].
Bubble custom fonts now work on mobile — Upload .ttf/.otf files directly instead of relying on web-safe fonts [15].

🔥 Top No-Code & AI News

1. OpenAI Previews GPT-5.6 Sol — Limited to Trusted Partners

OpenAI unveiled GPT-5.6 Sol on June 26, a next-generation model that advances coding, scientific reasoning, long-horizon planning, and agentic workflows [1]. The company also introduced two additional variants — Terra and Luna — rounding out the GPT-5.6 family [3].

What’s new in GPT-5.6 Sol:

Stronger cybersecurity capabilities — better at finding and fixing vulnerabilities than exploiting them, according to OpenAI’s system card [2].
Scientific reasoning gains — improved performance on research-oriented tasks and hypothesis generation.
Agentic workflow improvements — better long-horizon planning and multi-step task execution for AI agents.

The catch: OpenAI is limiting the rollout to “trusted partners” after the U.S. government requested restrictions on broad access. The company said restrictions “shouldn’t be the norm” in a statement accompanying the release [2][3].

Why it matters for no-code builders: GPT-5.6 Sol’s agentic workflow improvements will eventually reach tools that rely on OpenAI models — including Bubble’s AI Agent, Zapier AI, and Make AI Agents. The limited rollout means it’ll be a slower ramp than previous launches, so don’t expect it in your no-code tools right away.

2. U.S. Loosens Restrictions on Anthropic’s Mythos 5

In a parallel move, the Commerce Department gave Anthropic permission on June 26 to restore some clients’ access to Claude Mythos 5 — the company’s most capable frontier model [4]. Over 100 trusted U.S. organizations will regain access under revised licensing terms [5].

Mythos 5, which launched on June 9 alongside Claude Fable 5, was initially restricted under government safety review frameworks after Anthropic voluntarily disclosed certain capabilities [6].

Why it matters: Mythos 5 was Anthropic’s breakthrough in scientific hypothesis generation — in blinded tests, it outperformed human domain experts. For no-code builders, broader Mythos access means better reasoning capabilities in AI-powered automation tools that use Anthropic’s API. Expect platforms like Lindy, MindStudio, and n8n (which added a dedicated Anthropic node) to start offering Mythos-powered templates.

3. Google DeepMind’s Talent Exodus Accelerates

The week’s biggest story in AI talent: Google DeepMind is bleeding top researchers to rivals Anthropic and OpenAI.

The dominoes fell quickly:

Noam Shazeer — Google’s high-profile AI researcher who returned to the company in 2024 — left for OpenAI [7].
Jonas Adler and Alexander Pritzel — two senior DeepMind researchers — announced plans to join Anthropic [8][9].
Alphabet shares had their worst day in a year on June 22 as the market reacted [7].

DeepMind engineers are nearly 11 times more likely to leave for Anthropic than the reverse, according to a 2025 industry analysis [9].

Why it matters: When top talent flows from one AI lab to another, it signals where the technical edge is shifting. Anthropic is clearly building its bench, and Google’s AI dominance is showing cracks. For no-code builders, this could mean faster innovation cycles on Anthropic-powered tools and potentially slower progress on Google’s Gemini integrations in platforms like Make and n8n.

4. Zapier Moves AI Steps to Model-Based Pricing

Zapier changed how its AI features are priced on June 15, 2026. Instead of flat-rate AI step costs, AI by Zapier steps are now priced by model tier [11]:

Tier	Models	Task multiplier
Standard	GPT-4o mini, Claude Haiku, Gemini Flash	1x per step
Advanced (default)	GPT-4o, Claude Sonnet, Gemini Pro	3x per step
Premium	GPT-5.5, Claude Opus, Gemini Ultra	7x per step

This shift happened 48 hours after Anthropic changed its own Claude subscription billing (programmatic usage moved to a separate credit system), leading to what some outlets called a coordinated pricing pivot [12].

Why it matters: If you use Zapier’s AI steps for automation, your task consumption just changed dramatically. A workflow using GPT-4o (Advanced tier) now costs 3x per AI step compared to before. Review your Zaps and consider switching to Standard-tier models for non-critical tasks, or moving complex AI workflows to n8n or Make where pricing is more predictable.

5. Anthropic’s Claude Billing Change — Partially Delayed

Anthropic announced on May 14 that starting June 15, 2026, programmatic Claude usage (CLI, Agent SDK, GitHub Actions) would be moved off subscription plans and onto a separate credit system [13]. However, on June 16, Anthropic told subscribers the change is not taking effect yet — effectively pausing the rollout while gathering feedback [14].

Why it matters: This directly affects anyone using Claude Code, Claude’s Agent SDK, or Claude GitHub Actions as part of their no-code/AI workflow. The delay gives you more time to plan, but the direction is clear — Anthropic wants to separate API-level usage from chat subscriptions. If you rely on Claude’s programmatic access, budget for credits starting later this year.

6. Bubble: Custom Fonts on Mobile + Workflow Improvements

Bubble pushed several quality-of-life updates this week [15]:

Custom fonts now work on mobile — Upload .ttf or .otf files directly instead of relying on web-safe fonts. Previously, custom fonts only worked on desktop [15].
Workflow action disabling — You can now disable specific actions within a workflow without deleting them, making debugging and testing much easier [16].
Button + icon element improvements — Better styling control and icon alignment options [16].

Why it matters: The custom fonts on mobile update closes a long-standing gap for Bubble mobile apps. If you’ve been building mobile Bubble apps that needed branded typography, this eliminates a major workaround. The workflow action disable is a small change with big debugging impact — no more deleting and recreating actions to test conditional branches.

7. n8n: Continued AI Node Expansion

n8n has been steadily expanding its AI integration surface. Over 35 new nodes were added between January and June 2026, including dedicated nodes for [10]:

Anthropic Claude — direct integration without HTTP Request workarounds
Google Gemini — access to Google’s latest models
Perplexity — web-search-powered AI queries

n8n also shipped v2.28.0 on June 23 with bug fixes and stability improvements [17].

Why it matters: n8n’s aggressive node expansion makes it the most self-hostable AI workflow platform with first-class AI provider support. For no-code builders who need data privacy (GDPR, HIPAA), n8n’s self-hosted option with these native AI nodes is increasingly hard to beat.

⚡ Quick Hits

GPT-4.5 retirement — OpenAI announced GPT-4.5 will be retired from ChatGPT on June 26 (90-day sunset period). o3 retirement follows on August 26 [3].
Android 17 / June Pixel Drop — Google’s feature drop includes Gemini upgrades, floating app Bubbles for multitasking, and AI video/music creation tools. If you build no-code mobile tools, Android 17’s new API surface is worth a look [18].
Power BI June 2026 — Microsoft’s update includes Copilot-driven AI experiences and practical reporting improvements. Tighter AI integration in a tool many no-code builders use for dashboards [19].
Bubble’s Prompt AI Workflow Action — The simplified “talk to an LLM” step (no API Connector required) is still targeting end of June — expect it any day now [16].
Lindy pricing — Lindy updated its pricing page in June with plans starting at $69/month (not per user) [20].

📝 This Week on NoCode Insider

Here’s what we published this week:

Bubble vs FlutterFlow vs WeWeb (2026): Which No-Code Web App Builder Wins? (Jun 26) — Head-to-head comparison of the three leading visual app builders across pricing, performance, design flexibility, and AI features.
WeWeb Review 2026: The No-Code Builder That Developers Actually Like (Jun 25) — Hands-on review of WeWeb’s frontend builder, Xano integration, and AI agent features.
No-Code AI Agent Platforms 2026: The Complete Comparison (Jun 24) — Comprehensive comparison of 10+ no-code AI agent builders with real pricing, tested capabilities, and use case recommendations.
Build Your First AI Chatbot for Your Website (No-Code, 2026) (Jun 23) — Step-by-step guide to building and deploying an AI chatbot using no-code tools.
Activepieces vs n8n (2026): Which Open-Source Automation Tool Wins? (Jun 22) — Detailed comparison of n8n’s mature ecosystem vs Activepieces’ ease of use and growth trajectory.

🌐 From the Network

A few highlights from across the blog network this week:

NiteAgent: GPT-5.6 Sol First Look — Practical analysis of what Sol’s agentic workflow improvements mean for production AI agents, with benchmark numbers and migration guidance. (NiteAgent)
Hermes Tutorials: Deploying Multi-Provider Agent Workflows — A guide to routing AI tasks across DeepSeek V4 Flash, Claude Opus, and GPT-5.5 with fallback strategies and cost optimization. (Hermes Tutorials)
Code Intel: The DeepMind Exodus — Technical Impact Analysis — Engineering-focused breakdown of what the talent migration means for Google’s AI platform roadmap and downstream dependencies. (Code Intel)

📅 What’s Next

Bubble’s Prompt AI Workflow Action — the simplified LLM step without API Connector — ships any day now [16].
Anthropic’s Mythos 5 rollout to more organizations as the Commerce Department reviews expanded access [4].
OpenAI GPT-5.6 Sol — expect more trusted partners to announce integrations in the coming weeks [1].
Anthropic’s IPO roadshow continues — the confidential S-1 filing values the company at ~$965B [21].
n8n v2.28 stable is out — watch for v2.29 beta features in the coming weeks [17].

Sources

NiteAgent — AI agent development, frameworks, and production patterns
ToolBrain — tool reviews, LLM comparisons, and AI workflow guides
CodeIntel Log — code quality, debugging, and software engineering benchmarks
Hermes Tutorials — Hermes Agent setup, configuration, and advanced workflows

Cross-links automatically generated from NoCode Insider.