|
|
MODEL
SEISMIC
2026-06-26
GPT-5.6 — OpenAI previews Sol, Terra, and Luna tiers
OpenAI's new generation splits into three named tiers and adds an ultra mode that wires subagents into the flagship model.
What is it?
GPT-5.6 is OpenAI's next generation of general-purpose models, arriving roughly two months after GPT-5.5. It ships as three durable tiers — Sol for the hardest work, Terra as the everyday default, and Luna as the cheap and fast option.
How does it work?
Sol gets two new reasoning options: max effort for longer thinking on hard problems, and ultra mode that spins up subagents to tackle parts of a task in parallel. All three sizes show measurable jumps in coding, biology, and cybersecurity work over GPT-5.5.
Why does it matter?
The Sol/Terra/Luna split gives developers a clean way to choose between cost and depth with transparent day-one pricing ($5/$30 for Sol, $1/$6 for Luna per million tokens). For agent builders, ultra mode moves orchestration logic that used to sit in tools like LangGraph back inside the model itself.
Who is it for?
API users, agent builders, and ChatGPT power users. Access starts as a limited preview for trusted partners, with broader rollout in the coming weeks.
|
|
|
|
ECOSYSTEM
MAJOR
2026-06-26
Claude Mythos 5 restored — US Commerce lifts block for 100+ trusted partners
Commerce restores Claude Mythos 5 to 100+ Annex A trusted partners after two weeks dark — Fable 5 still under review.
What is it?
Claude Mythos 5, Anthropic's most capable model, is back online for 100+ trusted US institutions named in Commerce Secretary Howard Lutnick's Annex A letter. Two weeks ago the same office had ordered Anthropic to suspend Mythos 5 worldwide.
How does it work?
The arrangement is a narrow exemption from the original export-control directive. Annex A names 100+ US companies and federal agencies that Anthropic can serve without a separate deemed-export license. Foreign-national employees at those entities are covered as well.
Why does it matter?
For 14 days every paying customer was cut off worldwide. With the Annex A exemption, Fortune 500 firms and federal agencies on the list get Mythos 5 back, and Anthropic returns to revenue on the model. Fable 5 stays under review with no announced timeline.
Who is it for?
Fortune 500 partners and US federal agencies named in Annex A of the Commerce Secretary's letter.
|
|
|
|
REPO
MAJOR
2026-06-26
DSpark + DeepSpec — DeepSeek opens its speculative decoding stack
DeepSeek shipped a free codebase for training speculative-decoding drafters, plus DSpark drafters bolted onto V4-Pro and V4-Flash.
What is it?
DeepSpec is a full-stack MIT-licensed codebase from DeepSeek for training and evaluating draft models used in speculative decoding. The same release adds DSpark drafters as separate Hugging Face uploads that attach to existing V4-Pro and V4-Flash checkpoints instead of replacing them.
How does it work?
Speculative decoding pairs a fast draft model with the real target model — the drafter proposes several future tokens at once, then the target verifies them in one pass so accepted tokens skip ahead. DeepSpec packages the pipeline end-to-end with three drafter algorithms: DSpark, DFlash, and Eagle3.
Why does it matter?
Inference is where the bill is paid, and speculative decoding is the cheapest way to speed it up — but the draft model is the hard part to train. By open-sourcing both the training stack and the DSpark drafters for V4 checkpoints, DeepSeek lets self-hosters cut cost without changing the underlying model.
Who is it for?
Inference providers, self-hosters, and research teams running open-weight models who want faster token throughput.
|
|
|
|
MODEL
MAJOR
2026-06-25
Ornith 1.0 — open-weight coding models that learn their own RL scaffold
Open-weight agentic coding models that learn to write their own RL scaffold instead of relying on a fixed harness.
What is it?
Ornith 1.0 is an MIT-licensed family of agentic coding LLMs from DeepReinforce — 9B, 31B, 35B MoE, and 397B MoE variants built on Gemma 4 and Qwen 3.5 checkpoints — whose RL loop writes its own task-specific scaffold instead of using a fixed human-designed harness.
How does it work?
Each RL step runs in two stages: the model reads a coding task and proposes a refined scaffold, then uses that scaffold to generate a solution rollout. Across steps, the scaffold co-evolves with the model's policy — Ornith effectively learns the harness instead of inheriting a hand-built one.
Why does it matter?
Open coding models usually inherit a fixed human harness that caps self-improvement. Ornith ships an open recipe where the scaffold is a training target, and the 397B flagship hits 82.4% on SWE-Bench Verified — giving open-source teams a strong agentic coding baseline under MIT.
Who is it for?
Open-source agent builders and RL researchers who want a strong agentic coding baseline they can self-host and modify.
|
|
|
|
TOOL
MAJOR
2026-06-25
GitHub Copilot for Jira — GA brings agent streaming inside Atlassian
Run the GitHub coding agent on a Jira ticket, watch it stream, and steer it from Jira chat.
What is it?
GitHub Copilot for Jira is now generally available. The Atlassian Marketplace app links a Jira ticket to a GitHub coding agent session, so progress streams into Jira and follow-up instructions in the Jira chat keep the agent working on the same pull request.
How does it work?
Inside a Jira issue, the integration shows the agent's live steps as it runs and posts the resulting pull request back to the ticket. A reply in Jira chat continues the agent on the same PR. The app also pulls extra context from Confluence pages via MCP and supports custom agents and space-level guidance.
Why does it matter?
Project managers and developers who live in Jira can now drive code changes without switching tabs, see the agent's progress, and steer it without losing the existing pull request — cutting the context-switching between Jira and GitHub.
Who is it for?
Atlassian teams already using GitHub Copilot who want to drive code changes without leaving Jira.
|
|
|
|
RESOURCE
MAJOR
2026-06-26
Anthropic Economic Index: Cadences — Claude usage hour by hour
Anthropic's June 2026 Economic Index zooms in on how Claude usage rises and falls hour by hour.
What is it?
Cadences is the latest Anthropic Economic Index report — a recurring study of how people use Claude. This edition adds hourly sampling, a new artifact classifier sorting outputs into 30+ categories, and a survey linking what users say about AI to what they actually do.
How does it work?
Cadences blends hourly aggregated usage that exposes daily and weekly rhythms with an artifact classifier tagging outputs as code, documents, explanations, and 26 other types. It separately breaks out Claude.ai chat, the Cowork agent, and direct API traffic for the first time.
Why does it matter?
93% of conversations produce identifiable artifacts, higher-wage occupations consume ~2.5x more tokens than lower-wage ones, and over a third of surveyed users expect AI to handle most of their tasks within 12 months. Personal chats jump from ~35% of usage on weekdays to ~50% on weekends.
Who is it for?
Policy researchers, economists, AI labs, and journalists tracking how AI is actually being used at scale.
|
|
|
All releases at ai-tldr.dev
Simple explanations • No jargon • Updated daily
|
|
