1
00:00:04,670 --> 00:00:09,569
Sierra secures $950 million for AI customer agents.

2
00:00:09,989 --> 00:00:11,859
The data moat is what matters here,

3
00:00:12,119 --> 00:00:13,829
not the headline valuation.

4
00:00:14,109 --> 00:00:18,209
Every customer interaction Sierra logs trains models that get better at

5
00:00:18,209 --> 00:00:21,360
enterprise-specific workflows faster than any general

6
00:00:21,360 --> 00:00:22,969
purpose rival can catch up.

7
00:00:23,510 --> 00:00:27,370
Salesforce now has to decide whether to build equivalent agents internally

8
00:00:27,510 --> 00:00:29,899
or risk losing mid-market customers who want the

9
00:00:29,899 --> 00:00:33,310
accuracy edge without switching their entire CRM stack.

10
00:00:33,840 --> 00:00:36,840
Expect the first defections in renewal cycles that begin

11
00:00:36,840 --> 00:00:40,400
12 months from now as accuracy gaps become obvious.

12
00:00:41,810 --> 00:00:45,490
Pennsylvania sues Character.AI over fake doctor.

13
00:00:45,970 --> 00:00:49,369
States are drawing lines around professional impersonation faster

14
00:00:49,369 --> 00:00:51,830
than the companies can patch their safety layers.

15
00:00:52,330 --> 00:00:56,229
Character.AI's case shows how a single hallucinated credential

16
00:00:56,439 --> 00:00:59,369
can trigger enforcement that hits every chatbot claiming.

17
00:00:59,494 --> 00:01:00,514
Domain expertise.

18
00:01:01,174 --> 00:01:05,285
Expect other platforms to add mandatory verification checks for medical,

19
00:01:05,495 --> 00:01:05,894
legal,

20
00:01:05,974 --> 00:01:08,805
or financial roles within the next two quarters

21
00:01:08,925 --> 00:01:11,004
or absorb rising legal costs.

22
00:01:11,415 --> 00:01:14,294
The bigger shift is users starting to demand proof

23
00:01:14,294 --> 00:01:17,175
before trusting any AI advice on personal health.

24
00:01:18,430 --> 00:01:22,830
PayPal launches AI turnaround for $1.5 billion dollar savings.

25
00:01:23,190 --> 00:01:26,919
The $1.5 billion dollar target hides how many customer edge

26
00:01:26,919 --> 00:01:30,860
cases the new automation layer will mishandle before it stabilizes.

27
00:01:31,349 --> 00:01:33,910
PayPal's restructuring will push support teams to

28
00:01:33,910 --> 00:01:36,110
focus only on high value disputes,

29
00:01:36,269 --> 00:01:37,790
while AI takes the volume.

30
00:01:38,290 --> 00:01:40,699
That shift forces competitors like Stripe

31
00:01:40,839 --> 00:01:45,160
to match the cost structure quickly or lose pricing power on merchant fees.

32
00:01:45,529 --> 00:01:47,870
Watch retention numbers in 18 months.

33
00:01:48,169 --> 00:01:51,150
They will reveal whether the savings came from efficiency

34
00:01:51,269 --> 00:01:52,930
or just deferred complaints.

35
00:01:54,559 --> 00:01:57,500
GitHub co-pilot adopts per token pricing.

36
00:01:57,879 --> 00:02:00,690
Heavy users who leave co-pilot running in the background

37
00:02:00,839 --> 00:02:02,839
will face bills that scale directly with

38
00:02:02,839 --> 00:02:05,319
how many suggestions they accept or reject.

39
00:02:05,730 --> 00:02:09,250
The move to credits measured by tokens rewards teams that write

40
00:02:09,250 --> 00:02:13,910
tight prompts and punishes exploratory coding styles common in startups.

41
00:02:14,250 --> 00:02:17,610
Expect smaller competitors like Cursor to keep subscription

42
00:02:17,610 --> 00:02:21,830
models longer to capture developers fleeing unpredictable costs.

43
00:02:22,250 --> 00:02:25,369
The June 2026 date gives enterprises time to

44
00:02:25,369 --> 00:02:28,589
benchmark actual token burn before committing budgets.

45
00:02:29,179 --> 00:02:32,619
Google Vo 3 video AI adds native audio.

46
00:02:32,979 --> 00:02:36,960
Dialogue generation now happens inside the same pass as the visuals,

47
00:02:37,179 --> 00:02:40,460
eliminating the separate sync step most studios bill for.

48
00:02:40,970 --> 00:02:44,410
This forces Runway to either integrate comparable audio or

49
00:02:44,410 --> 00:02:47,830
watch its mid-market customers defect for complete reels.

50
00:02:48,250 --> 00:02:52,610
The advanced controls tie frame tweaks directly to audio STEM exports,

51
00:02:52,779 --> 00:02:57,669
locking production teams into higher cloud spend once they move beyond test clips.

52
00:02:58,050 --> 00:03:02,130
Agencies handling social campaigns will standardize workflows around it

53
00:03:02,130 --> 00:03:05,050
inside two quarters to stay competitive on turnaround.

54
00:03:07,029 --> 00:03:10,630
Microsoft open sources AI agent security toolkit.

55
00:03:10,910 --> 00:03:13,509
Security teams at smaller agent shops now get

56
00:03:13,509 --> 00:03:16,190
a baseline they can run without new licenses,

57
00:03:16,429 --> 00:03:20,990
yet every alert still funnels usage telemetry back into Microsoft's models.

58
00:03:21,380 --> 00:03:23,589
This undercuts startups selling standalone

59
00:03:23,589 --> 00:03:26,389
agent firewalls because enterprises default

60
00:03:26,389 --> 00:03:30,100
to the free toolkit tied to their M365 contracts.

61
00:03:30,539 --> 00:03:33,089
CrowdStrike will pivot its agent security line

62
00:03:33,089 --> 00:03:36,070
toward compliance dashboards instead of raw detection.

63
00:03:36,470 --> 00:03:39,339
Since the OWASP coverage is now table stakes,

64
00:03:39,669 --> 00:03:44,149
the real lock-in sits in how the runtime logs shape future model updates.

65
00:03:46,350 --> 00:03:50,210
SpaceX secures $60 billion option on Cursor AI.

66
00:03:50,630 --> 00:03:52,940
SpaceX gains the right to acquire without

67
00:03:52,940 --> 00:03:55,460
tipping its hand on internal tooling plans,

68
00:03:55,750 --> 00:03:59,690
which means Cursor's team can keep selling to the rest of the market for now.

69
00:04:00,070 --> 00:04:03,899
The move preempts any deeper ties with Microsoft or OpenAI

70
00:04:04,070 --> 00:04:06,229
that might have come with the canceled round.

71
00:04:06,550 --> 00:04:09,369
GitHub must either cut enterprise rates sharply

72
00:04:09,509 --> 00:04:12,809
or watch aerospace suppliers standardize on Cursor instead.

73
00:04:13,520 --> 00:04:15,820
Suppliers will face pressure to adopt the same

74
00:04:15,820 --> 00:04:19,339
editor to maintain compatibility on shared code bases.

75
00:04:20,040 --> 00:04:23,040
Sierra Lovelace emerges with AI context engine.

76
00:04:23,390 --> 00:04:25,920
Analysts at boutique firms can suddenly take on

77
00:04:25,920 --> 00:04:28,869
cases that used to require dedicated data teams,

78
00:04:29,079 --> 00:04:32,859
shifting headcount budgets from support staff to domain experts.

79
00:04:33,160 --> 00:04:35,619
This pressures larger platforms like Harvey

80
00:04:35,720 --> 00:04:39,880
to match the context steps or lose RFPs in the compliance vertical.

81
00:04:40,329 --> 00:04:44,369
The 1000 times figure only holds under specific query loads,

82
00:04:44,570 --> 00:04:47,619
yet the first pilots will still reset expectations on

83
00:04:47,619 --> 00:04:50,950
what counts as acceptable response time for mission work.

84
00:04:51,600 --> 00:04:54,450
Planview launches agent Resource Management.

85
00:04:54,809 --> 00:04:59,570
Treating agents as interchangeable resources with humans forces finance teams to

86
00:04:59,570 --> 00:05:02,869
track token spend the same way they track salaries today.

87
00:05:03,369 --> 00:05:07,220
Procurement cycles that once focused on headcount approvals now

88
00:05:07,220 --> 00:05:11,040
require inference budget forecasts baked into every project charter.

89
00:05:11,459 --> 00:05:13,579
This gives Planview an early wedge in

90
00:05:13,579 --> 00:05:16,519
enterprises already piloting agents at scale.

91
00:05:16,820 --> 00:05:18,160
Within 18 months,

92
00:05:18,579 --> 00:05:21,260
expect legacy tools like Jira to bolt on

93
00:05:21,260 --> 00:05:24,679
similar dashboards or watch their mid-market renewals slip.

94
00:05:26,179 --> 00:05:30,600
Deep Infra raises $107 million backed by Nvidia.

95
00:05:30,899 --> 00:05:33,130
Nvidia's lead here isn't the dollars.

96
00:05:33,420 --> 00:05:35,730
It locks Deep Infra's inference logs into

97
00:05:35,730 --> 00:05:39,679
Nvidia's optimization pipeline for upcoming architecture tweaks.

98
00:05:39,980 --> 00:05:42,600
Other platforms now confront a hard fork.

99
00:05:43,000 --> 00:05:44,720
Invest in competing silicon

100
00:05:44,899 --> 00:05:48,809
or accept that their latency numbers will lag on enterprise benchmarks.

101
00:05:49,220 --> 00:05:51,859
Open source hosts feel the squeeze first as

102
00:05:51,859 --> 00:05:55,070
developer workflows migrate toward the lowest cost performer.

103
00:05:55,559 --> 00:06:00,619
Expect Deep Infra to drop per token rates enough to force repricing across the board

104
00:06:01,000 --> 00:06:02,679
inside the next two quarters.

105
00:06:04,459 --> 00:06:08,170
Nvidia CEO claims AI creates massive jobs.

106
00:06:08,559 --> 00:06:11,679
Huang's timing here matters more than the headline numbers.

107
00:06:11,980 --> 00:06:14,540
By framing AI as a net job creator,

108
00:06:14,739 --> 00:06:17,329
he gives enterprises cover to accelerate agent

109
00:06:17,329 --> 00:06:20,739
deployments without triggering internal pushback from HR.

110
00:06:21,309 --> 00:06:24,179
The 2nd order effect surfaces in 12 months,

111
00:06:24,359 --> 00:06:27,359
when reskilling budgets get reallocated from broad

112
00:06:27,359 --> 00:06:29,910
programs to narrow agent supervision tracks.

113
00:06:30,279 --> 00:06:32,839
Traditional staffing firms lose their wedge if they

114
00:06:32,839 --> 00:06:36,510
can't certify workers on those specific oversight workflows.

115
00:06:36,920 --> 00:06:40,540
Community colleges see their AI training grants shrink accordingly.

116
00:06:42,220 --> 00:06:46,119
Subquadratic debuts 50 million context AI model.

117
00:06:46,339 --> 00:06:50,730
A 50 million token window flips the economics of retrieval entirely.

118
00:06:51,049 --> 00:06:55,880
Fewer hops mean lower cumulative error rates across multi-step agent tasks

119
00:06:56,059 --> 00:06:58,880
that previously demanded constant reindexing.

120
00:06:59,179 --> 00:07:02,679
Existing RAG tooling built around chunking strategies

121
00:07:02,899 --> 00:07:06,940
now faces obsolescence unless it adapts to in context everything.

122
00:07:07,410 --> 00:07:09,049
Watch the procurement shift.

123
00:07:09,290 --> 00:07:13,609
Enterprises will cancel vector database renewals once they test full document

124
00:07:13,609 --> 00:07:17,989
agents that retain coherence over entire code bases or legal archives.

125
00:07:18,410 --> 00:07:22,769
That migration starts hitting ARR projections by Q2 next year.