AI Tech News

2 / 20

AI Research

AI Research This Week: Autoregressive To Diffusion, Mixture-of-Experts Drafting

Research this week explores converting autoregressive models to diffusion language models, speculative decoding for mixture-of-experts, and how agent context pipelines lose information.

15m ago

3 / 20

Artificial Intelligence

Build a factory, not just a draft

Zack Proser says the danger is not AI slop but untold stories, and shows a Slack factory that gets drafts 85 to 90 percent ready.

7h ago

Apollo Says AI Has Paid Zero Outside Tech So Far

4 / 20

Market Research

Apollo Says AI Has Paid Zero Outside Tech So Far

Apollo's Torsten Slok finds AI has delivered zero financial payoff for the S&P 493, with flat margins and revenue per worker outside seven megacaps.

8h ago

Nick Nisi's Fruit Fly Code Review Is Real

5 / 20

Artificial Intelligence

Nick Nisi's Fruit Fly Code Review Is Real

WorkOS engineer Nick Nisi wired a fruit fly connectome to an AI nose and let it vote on PRs to argue human code review must change.

12h ago

Alex Hormozi says Change My Mind and Win $10000

6 / 20

Startup News

Alex Hormozi says Change My Mind and Win $10000

Alex Hormozi offered $10,000 to anyone who could change his mind, then conceded on work-life balance while holding his line on A players, pricing and weekends.

14h ago

Vercel Wants One Person And Agents To Build $1T

7 / 20

Artificial Intelligence

Vercel Wants One Person And Agents To Build $1T

Vercel CEO Guillermo Rauch told Delphi his platform could host a single-person trillion-dollar company, with AI agents doing the work of thousands.

14h ago

Pretraining Poison Is Now a Remote Code Bug

8 / 20

AI Research

Pretraining Poison Is Now a Remote Code Bug

Pretrained Qwen3 agents hit 92.7% remote code execution from 0.1% pretraining poison that survived clean SFT, DPO and GRPO.

21h ago

Compound Rebuilt Excel So Agents Can Run It

9 / 20

Artificial Intelligence

Compound Rebuilt Excel So Agents Can Run It

Compound AI rebuilt Excel, PowerPoint and Word as headless Rust engines so buy-side agents can run in parallel with cited, inbox-native teamwork.

22h ago

Mintlify says 80% of docs traffic is now agents

10 / 20

Artificial Intelligence

Mintlify says 80% of docs traffic is now agents

Mintlify CTO Hahnbee Lee said AI agents now drive 80% of docs views, pushing Coinbase to expand x402 micropayments for agents.

1d ago1

11 / 20

Artificial Intelligence

If You Score 100%, Your Test Failed

Ramakrishnan Lokanathan told AI Engineer a 100% eval means the test is wrong, and showed E-Slams, a game-based harness for agents.

Linux TCP flaw fixed upstream, not in Ubuntu or RHEL

12 / 20

Artificial Intelligence

Linux TCP flaw fixed upstream, not in Ubuntu or RHEL

Atlas says CVE-2026-98432 got a four-line upstream fix Oct 15, but Ubuntu and RHEL kernels remain vulnerable as of Oct 31.

Oracle Shows Single MCP Gate to Lock Agent Data

13 / 20

Technology

Oracle Shows Single MCP Gate to Lock Agent Data

Oracle's Jeff Smith showed on AI Engineer how a single MCP server with governed reports and deep data security can give agents data without leaking PII.

Oracle says memory, not models, is the moat

14 / 20

Technology

Oracle says memory, not models, is the moat

Oracle's Richmond Alake demoed agent memory and governance on AI Engineer, pitching owned memory and identity-bound access as the differentiator.

1d ago2

15 / 20

Artificial Intelligence

Musk says AI will erase money and work

Musk told The Builders that superintelligence could add $30T a year and make money irrelevant, with robots mattering more than LLMs.

Bumble found 78% wanted AI photo feedback but only 10% used it

16 / 20

Artificial Intelligence

Bumble found 78% wanted AI photo feedback but only 10% used it

Bumble's AI photo feedback won 78% support in research but drew 10 to 15% usage in its AB test, forcing guardrails and cheaper models.

2d ago1

IMEC shows when owning AI models beats API bills

17 / 20

Artificial Intelligence

IMEC shows when owning AI models beats API bills

IMEC's Steven Latré told AI Engineer when hosting open-weight models beats APIs: GLM-class coding workloads saved 24-25% and Kimi K3 hit 86% but only four concurrent users.

2d ago1

One Wrong Question Breaks the Voice Agent

18 / 20

Artificial Intelligence

One Wrong Question Breaks the Voice Agent

Upstart showed how a verification call fails on one illegal follow-up and how structured evals lifted judge-human agreement from 33% to 92%.

2d ago1

Northflank says assume all secrets were decrypted

19 / 20

Artificial Intelligence

Northflank says assume all secrets were decrypted

Northflank told customers to assume all environment variables were decrypted and rotate every secret after reports of abused AI API keys.

2d ago

Oracle Lays Out When AI Agents Should Remember

20 / 20

Artificial Intelligence

Oracle Lays Out When AI Agents Should Remember

Oracle's Wojciech Pluta detailed a memory promotion gate on AI Engineer to decide what agents keep, with RAG as the read path and memory as the write path.

2d ago

AI writes faster than humans can review

Keep scrolling for more

Next page loads automatically as you reach this card.

Technology

Sourcegraph's Andrew Norrish told AI Engineer that review is now the bottleneck as AI-written PRs double in count and size.

Daniel Singer• 5 min read•Nov 2 at 9:17 AM

Latest

Artificial Intelligence

Nick Nisi's Fruit Fly Code Review Is Real

WorkOS engineer Nick Nisi wired a fruit fly connectome to an AI nose and let it vote on PRs to argue human code review must change.

Alex Hormozi says Change My Mind and Win $10000

Startup News

Alex Hormozi offered $10,000 to anyone who could change his mind, then conceded on work-life balance while holding his line on A players, pricing and weekends.

Vercel Wants One Person And Agents To Build $1T

Artificial Intelligence

Vercel CEO Guillermo Rauch told Delphi his platform could host a single-person trillion-dollar company, with AI agents doing the work of thousands.

Pretraining Poison Is Now a Remote Code Bug

AI Research

Pretrained Qwen3 agents hit 92.7% remote code execution from 0.1% pretraining poison that survived clean SFT, DPO and GRPO.

Compound Rebuilt Excel So Agents Can Run It

Artificial Intelligence

Compound AI rebuilt Excel, PowerPoint and Word as headless Rust engines so buy-side agents can run in parallel with cited, inbox-native teamwork.

Mintlify says 80% of docs traffic is now agents

Artificial Intelligence

Mintlify CTO Hahnbee Lee said AI agents now drive 80% of docs views, pushing Coinbase to expand x402 micropayments for agents.

If You Score 100%, Your Test Failed

Artificial Intelligence

Ramakrishnan Lokanathan told AI Engineer a 100% eval means the test is wrong, and showed E-Slams, a game-based harness for agents.

Linux TCP flaw fixed upstream, not in Ubuntu or RHEL

Artificial Intelligence

Atlas says CVE-2026-98432 got a four-line upstream fix Oct 15, but Ubuntu and RHEL kernels remain vulnerable as of Oct 31.

5 min read•1d ago

Technology

Oracle Shows Single MCP Gate to Lock Agent Data

Oracle's Jeff Smith showed on AI Engineer how a single MCP server with governed reports and deep data security can give agents data without leaking PII.

Oracle says memory, not models, is the moat

Technology

Oracle's Richmond Alake demoed agent memory and governance on AI Engineer, pitching owned memory and identity-bound access as the differentiator.

Musk says AI will erase money and work

Artificial Intelligence

Musk told The Builders that superintelligence could add $30T a year and make money irrelevant, with robots mattering more than LLMs.

Bumble found 78% wanted AI photo feedback but only 10% used it

Artificial Intelligence

Bumble's AI photo feedback won 78% support in research but drew 10 to 15% usage in its AB test, forcing guardrails and cheaper models.

IMEC shows when owning AI models beats API bills

Artificial Intelligence

IMEC's Steven Latré told AI Engineer when hosting open-weight models beats APIs: GLM-class coding workloads saved 24-25% and Kimi K3 hit 86% but only four concurrent users.

One Wrong Question Breaks the Voice Agent

Artificial Intelligence

Upstart showed how a verification call fails on one illegal follow-up and how structured evals lifted judge-human agreement from 33% to 92%.

Northflank says assume all secrets were decrypted

Artificial Intelligence

Northflank told customers to assume all environment variables were decrypted and rotate every secret after reports of abused AI API keys.

5 min read•2d ago

Artificial Intelligence

Oracle Lays Out When AI Agents Should Remember

Oracle's Wojciech Pluta detailed a memory promotion gate on AI Engineer to decide what agents keep, with RAG as the read path and memory as the write path.