In partnership with |
 |
|
|
|
| |
| |
| |
Good morning, AI enthusiast. xAI's Grok Bot just graduated from a tool you assign tasks to into a teammate that spots work on its own, joins your Slack and Teams channels, and follows up on what's left undone. |
Shared memory, video-meeting attendance and permission-aware approvals are all on the table. Will teams actually hand an AI the chance to nudge them, or is this one more notification to mute? |
Today in AI Brief: |
Grok Bot becomes a proactive teammate DeepMind's AlphaGenome 2 reads DNA by cell type Anthropic's guide to choosing a Claude |
|
|
|
|
|
| |
|
The AEO work, without the AEO hire. |
|
Most AI visibility tools show you where you rank, then leave the work to you. |
ChatFeatured is an AI search agent that does it: finds the questions competitors win, drafts the content in your voice, and tracks the lift. |
Your team decides what ships. Month to month. |
Try the agent, today. |
| |
| |
| |
xAI Turns Grok Bot Into a Proactive Team Member |
In Brief: xAI released Grok Bot 1.0, a major update that lets shared bots spot work that needs doing, offer to take it on, and follow up on unfinished tasks. Bots now live in Slack, Microsoft Teams and video calls. |
The Details: |
Bots can be shared with people or groups and learn from everyone they work with, while escalating decisions that need a human's judgment. Safeguards include respecting each user's permissions, asking before sensitive actions and checking whether information suits a public channel, plus a "Smarter approvals" update that cuts needless confirmation prompts. Team Bots require a Team or Enterprise plan, and meeting attendance needs Team, Enterprise or Ultra, with Enterprise admins switching on Meetings in settings. |
Take Away: |
AI assistants are shifting from waiting for prompts to watching the work queue. The test is whether teams trust a bot enough to let it interrupt them, and the permission controls are what will decide that. |
|
|
|
|
|
| |
|
| |
| |
| |
DeepMind's AlphaGenome 2 Predicts DNA Effects Cell by Cell |
In Brief: Google DeepMind announced AlphaGenome 2, a model that predicts how genetic variants change gene activity in specific human cell types, rather than averaging across whole tissues. |
The Details: |
On a dataset of 28 immune cell types, it identified causal variants in 30% of cases, up from 16% for the original fine-tuned model. It was trained on more than 100 million individual cell profiles across 68 tissues, and takes a 1-million-letter DNA sequence as input. In inflammatory bowel disease, it reproduced a known finding involving the ETS2 gene in gut macrophages, and it can adapt to new single-cell data without retraining. |
Take Away: |
Knowing which cells a DNA change affects helps researchers decide which experiments to run first. It is available for non-commercial research through the AlphaGenome API, and DeepMind says it has not been validated for clinical use. |
|
|
|
|
|
| |
|
|
Analytics on Live Data Without Leaving Postgres |
|
When analytics on Postgres slows down, most teams add a second database. Then come the pipelines, the sync jobs, and a copy of your data that's always a little behind. |
TimescaleDB takes a different approach: extend Postgres instead of splitting away from it. Hypertables partition your data automatically as volume grows. Hypercore compression cuts storage up to 95%. Continuous aggregates keep dashboards live without re-querying everything. |
CERN runs Postgres this way for sensor data from the Large Hadron Collider. |
No split architecture, no pipeline lag, no new query language to learn. Same SQL, same drivers, same tools. |
Start on Tiger Cloud and get $1000 in credits. |
Get $1000 Credit |
| |
| |
| |
Anthropic's Guide Tells You Which Claude to Use |
In Brief: Anthropic's developer site published a guide to choosing a Claude model, recommending most people start on Opus 5.5 at medium effort and move up or down by task. |
The Details: |
Opus 5.5 is the daily driver for coding and knowledge work, Sonnet 5.5 suits fast, well-scoped jobs, and Haiku 5.5 handles classification and light subagent work. Fable 5.2 is pitched for major refactors, architecture planning and leading many subagents, and the guide reports it at high effort roughly matching Opus 5.5 at max on SWE-bench Pro (45.3% vs. 45.5%) at about half the cost per task. When a model falls short, the guide's order is add a test or script, raise effort one level, then move up a model, and it includes a ready-to-copy Haiku subagent for reading logs. |
Take Away: |
Picking the right model and effort level now changes both your bill and your results. Copy the log-reader subagent from the guide and run your own tasks through two models to see where the money is going. |
|
|
|
|
|
| |
|
| |
| |
| |
Everything else in AI |
Anthropic's upcoming IPO has reportedly drawn orders several times the shares on offer, according to Bloomberg. |
OpenAI posted its first $10B quarter in Q3, per Bloomberg. |
United Airlines is blocking unauthorized AI agents from its platforms and outside booking sites, citing accidental purchases and rules requiring human sign-off. |
Google introduced the Gemini Decisions API for fast yes/no answers. |
Blue opened orders for its AI phone beta, a Pixel 10 Pro running Blue OS at $1,299 that drops to $0 with an iPhone trade-in. |
|
|
|
|
|
| |
|
|
|
