Daily AI intelligence for Iru.
Monday, August 24, 2026
  • OpenAI launches Ultrafast mode for GPT-5.6 Sol at 14x speed with a 20%-plus API price cut.
  • Salesforce ships Slack Code, putting AI coding agents into shared team channels across all plans.
  • Anthropic commits $35M in Claude credits to open-source vulnerability remediation via its Defender Advantage Fund.
  • Nvidia notifies major customers of 15%-plus price hikes on AI servers including Vera Rubin and Grace Blackwell.
  • Claude Code overtakes GitHub Copilot as the most-used AI coding tool among professional developers, per JetBrains.
OpenAI Ships Ultrafast GPT-5.6 Sol, Cuts API Price
cio.economictimes.indiatimes.com · Aug 24

OpenAI introduced an Ultrafast mode for GPT-5.6 Sol delivering response times up to 14x faster, paired with a 20%-plus developer price cut for the next three months.

  • Speed target is agents: the latency reduction is explicitly positioned for real-time agentic workflows where sequential step delays compound.
  • Price cut is tactical: the three-month window on the discount signals competitive pressure from Anthropic and Google rather than a permanent margin concession.
Bottom line
OpenAI is using simultaneous speed and price moves to lock in enterprise agent workloads before competitors can respond with their next frontier release.
Salesforce Ships Slack Code for Team-Visible Agent Coding
newsbytes.ph · Aug 24

Salesforce launched Slack Code, dedicated team channels where AI coding agents work in the open with four tabs — Conversation, Plan, Code Diffs, and Live Preview — visible to the whole team.

  • Available across all Slack plans: supports Claude Code, Devin, GitHub Copilot, OpenAI ChatGPT, and Vercel agents, though each requires a separate subscription.
  • Anyone can redirect the agent mid-task: the architecture moves AI-assisted development from a solo, private-tab workflow to a team-steered, auditable process.
Bottom line
Slack Code is less a coding tool than a governance layer — by making agent actions visible in channels, Salesforce is betting that team oversight becomes a purchasing criterion for enterprise software dev.
Anthropic Expands Mythos 5 Access, Funds Open-Source Security
cryptobriefing.com · Aug 24

Anthropic is bringing Claude Mythos 5, its restricted frontier cybersecurity model, into security vendor products and committing $35M in Claude credits through its Defender Advantage Fund to fix vulnerabilities in widely used open-source software.

  • Claude Security scans are first: Claude Enterprise customers get Mythos 5-powered scans immediately, with broader cybersecurity vendor integrations planned next.
  • $35M is credits, not cash: the fund targets live vulnerabilities in high-impact open-source projects and funds automated patching approaches designed to eliminate whole classes of attacks.
Bottom line
Anthropic is using Mythos 5 as a wedge into the security vendor stack while simultaneously building goodwill with the open-source community that underlies most enterprise infrastructure.
Claude Code Tops JetBrains Developer Survey Over Copilot
digitaltoday.co.kr · Aug 24

JetBrains' Developer Ecosystem Survey 2026 finds 90% of professional developers use AI coding agents weekly, with Claude Code now the most-used tool, displacing long-time leader GitHub Copilot.

  • Workplace adoption is the story: the survey specifically measures professional, on-the-job use, not hobbyist experimentation, making the Copilot displacement commercially significant.
  • Agent-class tools dominate: the survey marks a shift where coding agents rather than inline autocomplete tools define the category.
Bottom line
GitHub's distribution advantage through VS Code did not hold — Claude Code's agent-first architecture proved more compelling to working developers than Copilot's completion-centric model.

Nvidia is in talks to invest in Perplexity at a valuation above $30B — more than 50% above its last round — with a technology licensing arrangement also under discussion, signaling that chip vendors are buying stakes in inference-heavy AI products to guarantee demand for their hardware.

Databricks, crossing $7B in annualized revenue at 80%-plus growth, reports its serverless Postgres product Lakebase reached $100M ARR in year one, with a fresh $5B funding round at a $190B valuation earmarked for Lakebase, Genie, and Unity AI Gateway — a signal that agent-native data infrastructure is a standalone market, not a feature.

Meta has hired OpenAI researcher Luke Metz for its Superintelligence Labs starting this week, continuing a talent acquisition spree that has already included a $14.3B Scale AI stake and triple-digit compensation packages — suggesting Meta is treating researcher headcount as the primary input to close the gap with frontier labs.

SentinelOne Earnings Show AI-Fueled ARR Growth Accelerating
tipranks.com · Aug 24

SentinelOne's latest earnings call highlighted AI-driven product adoption as the primary growth lever, with Q1 FY27 EPS of $0.04 beating the $0.02 estimate by 80% and the company projecting 22% year-over-year revenue growth for Q2.

  • Purple AI is the named driver: management attributed ARR acceleration directly to Purple AI adoption expanding across the customer base following the Black Hat open-access move.
  • Beat on profit, miss on revenue: Q1 revenue of $276.7M came in 2.2% below the $282.9M consensus, suggesting AI upsell is improving margins faster than it is adding net-new logos.
Bottom line
SentinelOne's results suggest the security AI upsell thesis is working on profitability first and top-line growth second — a sequencing that matters for how competitors price their own AI tiers.
  • Zero model loyalty among developers Developers on X are sharing charts showing DeepSeek v4 Flash usage obliterated by a new open-weight competitor in under 72 hours, sparking debate about whether any AI lab has a durable user base or whether the market is permanently switching-cost-free.
  • OpenAI gaining on Anthropic with enterprises Multiple accounts are circulating data showing OpenAI closing the enterprise ARR gap with Anthropic as businesses swap models freely on new releases, prompting concern that the market's fluidity makes lab revenue projections unreliable.
  • Small models plus agents beating large models Builders are sharing benchmarks where 27B-parameter models paired with purpose-built research agents outperform much larger monolithic models on specific tasks, fueling a broader argument that architecture beats scale at the task level.
  • LLM agent credit assignment failures A weekend paper showing current credit-assignment methods cannot reliably identify which agent steps caused an outcome is getting traction among practitioners who say it explains why their multi-step agents are hard to debug and improve.
  • AI bots drowning niche X conversations Ethan Mollick and others are complaining that well-trained LLM bots now reply coherently to obscure academic references on X, making it impossible to tell genuine expert engagement from automated response — a sign that social signal quality on the platform is degrading for researchers.

← Back to latest