Skip to content
LiveNext 12:41:43
topfive.ai
Today's five/N°13 · Sunday, October 18, 2026/Filed 04:50 ET

AI news today: DeepSeek opens a search API, Anthropic hires fewer juniors

One cheaper model, one narrower first rung, and three reminders that whoever holds your defaults, your keys and your terms holds a piece of your work.

The five · ranked by consequence

Permalink ↗
  1. ModelsOpen source3 min

    DeepSeek updates V4.1 Flash and opens its search API

    DeepSeek released V4.1-Flash-1017 under the MIT license on Oct. 17, and says it scores 77.0 on DeepSWE v1.1, up from 74.2, at an unchanged API price. The model is a modest step. The $0.75-per-1,000 search API, free in Claude Code and Codex until Nov. 1, is the part to try.

    Why it matters If you run agents on a budget, you now get an open model DeepSeek says is cheaper per task and a web search tool from the same provider, with free search calls until Nov. 1. Test both on your own tasks this week, and check where your queries go before customer data touches either.

    SRCDeepSeek API Docs

  2. BusinessAgents3 min

    Anthropic's junior tech hiring falls to 9%, from 27%

    Anthropic says junior hires fell from 27% of its technical hiring to 9%, while median meeting time rose 20% in a year, in a report on its own workforce published Oct. 16. The bottleneck has moved to human attention, and the company selling the tool says so.

    Why it matters If you manage people who work with agents, plan their time for briefing and reviewing, not typing. If you're early in your career, the lab with every reason to hire is taking fewer juniors into technical roles, so show interviewers how well you use AI tools.

    SRCAnthropic

  3. SafetyAgents2 min

    OpenCode breach puts stored OpenAI and Anthropic keys at risk

    OpenCode says a breach tied to a "large scale distillation attack" hit 1.4% of OpenCode Console users, all of whom had linked outside model providers. If you saved a provider key there, rotate it today and check your bill.

    Why it matters A stolen model key spends your money on someone else's training run. If you got OpenCode's email, revoke and reissue those keys at the provider today, then set spending limits on every account and count how many other tools hold one of your keys.

    SRCOpenCode

  4. ProductsBusiness2 min

    Replit makes OpenAI the default for Replit Agent

    Replit CEO Amjad Masad says OpenAI is now the preferred partner for Replit Agent, citing cost, partnership and performance, though Claude stays in the model selector. Most users never change the default, so the default is the decision.

    Why it matters If you build on Replit and never touched the model selector, the model writing your app has likely changed. Rerun the flows you depend on, and pick Claude yourself if it did better for your work. If you build on someone else's models, keep your own evals: partners move.

    SRCTechCrunch

  5. PolicyBusiness2 min

    Meta drops AI training language from its business terms

    Meta removed a clause from its updated commercial terms that would have let it use content from sites tied to business accounts to train its products, TechCrunch reports. The crawling permission stays, and it takes effect Oct. 30.

    Why it matters If your domain is linked to a Meta business or ads account, you accept these terms, and from Oct. 30 they cover Meta's crawlers on your pages. The training clause is gone. Meta says robots.txt still lets you manage access, so make yours say what you mean before the date.

    SRCMeta

Prompt of the day

Audit every AI API key you've handed to a tool

› I want to audit where my AI API keys live, after a coding tool's breach put stored provider keys at risk. Here is my setup:

- Providers I pay for: [e.g. OpenAI, Anthropic, Google]
- Tools, apps and services I've pasted a key into: [coding agents, IDE plugins, gateways, no-code tools, CI jobs]
- Where keys sit on my machines: [.env files, config files, password manager]

Do four things:
1. Make a table: each place a key lives, which provider it belongs to, who else can see it, and what a breach of that tool would cost me (spend, data, or both).
2. Rank them by risk and tell me which three to rotate first, and why.
3. For each provider, list the steps to revoke a key, issue a new one, set a monthly spending limit and review recent usage. If you're not sure of a provider's exact menu names, say so instead of guessing.
4. Suggest ways to cut the number of places my keys live (one key per tool, scoped keys, a single gateway), with the trade-off of each.

Ask me one question first if anything in my list is unclear.
Open in Magai

Works in any AI chat. In Magai, run it against several models at once.

Most in the news · 14 days

Every day, on the record.

Each square is an edition. Click any day to read the five that mattered then.

Editions
13
Stories ranked
65
Topics tracked
10
Terms defined
57
AprJunJulAugSepOct
EditionNo editionTodayFull archive →

By topic

All topics →

Models

Agents

Research

Business

Policy

Products

Open source

Hardware

Creative

Safety

Speak fluent AI.

Every term our stories use, defined in plain English and linked wherever it appears.

The glossary →