Latest / The Augur Dispatch / Barclays' Claude Code rollout: price the pilot before expanding access

Transcript

Barclays' Claude Code rollout: price the pilot before expanding access · The Augur Dispatch

• 0:06
  SPEAKER_00: A coding tool can become popular before anyone knows what a full rollout will cost.
• 0:11
  SPEAKER_00: Barclays expects half its developers to use Claude Code by the end of this year.
• 0:16
  SPEAKER_00: Before you approve a target like that, price the active users in your pilot and decide what result would justify expanding it.
• 0:22
  SPEAKER_00: This is the Augur dispatch for Friday, October 2nd, 2026.
• 0:27
  SPEAKER_00: The adoption case is strong.
• 0:29
  SPEAKER_00: Anthropic says Barclays expects Claude Code to reach half its developers this year, then a majority of software engineers in 2027.
• 0:37
  SPEAKER_00: Those are forecasts, not completed milestones, and both refer to the coding tool.
• 0:42
  SPEAKER_00: The uncomfortable comparison comes from Uber.
• 0:45
  SPEAKER_00: In a conference talk, Klein's Saoud Rizwan relayed an account attributed to Uber's chief technology officer.
• 0:52
  SPEAKER_00: Very heavy clod use, and a 2026 budget exhausted in four months.
• 0:57
  SPEAKER_00: The account doesn't identify which budget.
• 1:00
  SPEAKER_00: It isn't a statement about Uber's total spending, and it doesn't tell us whether the work produced was worth the money.
• 1:06
  SPEAKER_00: But it gives an executive a sensible reason to ask for the bill before approving the next wave of access.
• 1:13
  SPEAKER_00: Picture an engineering manager preparing next quarter's request.
• 1:17
  SPEAKER_00: The pilot is popular, developers want more access, and the easiest number to put on the slide is how many people will get it.
• 1:24
  SPEAKER_00: That slide can be completely accurate and still leave finance unable to answer the question it was brought into the meeting to answer.
• 1:31
  SPEAKER_00: The Uber Conference account describes roughly 5,000 engineers getting access in December last year.
• 1:37
  SPEAKER_00: It says 95% of engineers used Claude and 70% of committed code came from it.
• 1:44
  SPEAKER_00: Committed code means code saved into the team's shared working record.
• 1:48
  SPEAKER_00: Those numbers describe adoption and activity.
• 1:51
  SPEAKER_00: They don't establish that 70% of useful work was automated, or that the resulting software was reliable, or that review became cheaper.
• 2:00
  SPEAKER_00: Barclays' announcement offers no measured cost or quality result for this coding rollout.
• 2:05
  SPEAKER_00: It does describe other uses.
• 2:07
  SPEAKER_00: A knowledge assistant adopted by more than 16,000 colleagues, and a system processing roughly 120,000 emails a day.
• 2:15
  SPEAKER_00: We shouldn't turn missing coding economics into a claim that Barclays has published nothing about its wider use of AI.
• 2:22
  SPEAKER_00: Equally, those other workflows don't tell a software director what clawed code will cost per engineer.
• 2:28
  SPEAKER_00: A regulated bank's adoption is evidence of willingness to deploy.
• 2:32
  SPEAKER_00: It isn't an audit of the return your own organization will earn, with your systems, your contracts, and your review standards.
• 2:40
  SPEAKER_00: The first useful check doesn't require a new dashboard.
• 2:44
  SPEAKER_00: Take last month's pilot bill and divide it by the engineers who actually use the tool, rather than everyone who had a license.
• 2:51
  SPEAKER_00: Multiply that average by the group you plan to cover.
• 2:54
  SPEAKER_00: Then, compare the projected monthly bill with the approved budget and the months it needs to last.
• 3:00
  SPEAKER_00: This is a planning estimate, not a quote.
• 3:02
  SPEAKER_00: It assumes the wider group will use the tool at roughly the pilot's intensity, and contract discounts or heavier workloads could change the result.
• 3:11
  SPEAKER_00: Test a busier month too.
• 3:13
  SPEAKER_00: A spreadsheet is allowed to imagine enthusiasm.
• 3:16
  SPEAKER_00: Keep the generated share of code beside the bill as an activity measure, but add an outcome that matters.
• 3:22
  SPEAKER_00: Reviewed changes accepted without extra defects, for example.
• 3:26
  SPEAKER_00: Imagine a developer getting a first draft finished earlier, only to spend the afternoon correcting it with a reviewer.
• 3:33
  SPEAKER_00: That could still be a useful tool, but counting the first draft alone would miss part of the work you paid for.
• 3:39
  SPEAKER_00: Auger's recommendation is a bounded pilot on one repeatable workflow, such as routine maintenance changes.
• 3:46
  SPEAKER_00: Have the engineering lead own it with a finance partner.
• 3:49
  SPEAKER_00: Compare with the team's recent baseline, including review effort and defects, and agree on a review after one month.
• 3:56
  SPEAKER_00: Pause expansion if spending exceeds the agreed allowance or quality worsens.
• 4:01
  SPEAKER_00: Expand only when the team can show useful, accepted work at a cost it can sustain.
• 4:06
  SPEAKER_00: Access alone isn't much of an advantage when competitors can buy the same tool.
• 4:11
  SPEAKER_00: Learning which changes it handles reliably in your code base and improving the review process around those changes could be.
• 4:19
  SPEAKER_00: Model pricing adds another reason to leave room to change course.
• 4:23
  SPEAKER_00: Google has announced Gemini 4 Argonne with an introductory price of$2 per million input tokens and$10 per million output tokens.
• 4:32
  SPEAKER_00: Tokens are the pieces of text a model reads and writes.
• 4:36
  SPEAKER_00: That matches the standard input and output prices OpenAI publishes for GPT 6.1 Sol.
• 4:42
  SPEAKER_00: Google is positioning Argonne as a frontier model.
• 4:45
  SPEAKER_00: OpenAI describes Sol as approaching Astra's capabilities at a lower price.
• 4:50
  SPEAKER_00: But the discount has a stated destination.
• 4:53
  SPEAKER_00: Google's own footnote says the prices will become$4 for input and$20 for output after the introductory period.
• 5:00
  SPEAKER_00: The end date remains unspecified in the announcement.
• 5:04
  SPEAKER_00: Access matters too.
• 5:05
  SPEAKER_00: Argon is rolling out to trusted testers and cyber defenders through Google's Fairwind program.
• 5:11
  SPEAKER_00: The announcement describes broader release as still to come, so an ordinary buyer can't treat it as an immediately available replacement.
• 5:18
  SPEAKER_00: There is outside testing beyond the Valse ranking.
• 5:21
  SPEAKER_00: Artificial Analysis has published an evaluation, including a cost comparison that shows why matching token prices doesn't guarantee matching task costs.
• 5:31
  SPEAKER_00: Different models can use very different amounts of text to finish the same work.
• 5:36
  SPEAKER_00: Picture the procurement lead putting the introductory rate into next year's budget.
• 5:40
  SPEAKER_00: The correction belongs in that spreadsheet now, while changing a cell is still the whole job.
• 5:46
  SPEAKER_00: OpenAI's separate decision to shelve the planned release of GPT 6.1 Astra also needs precise wording.
• 5:53
  SPEAKER_00: It was a release decision, not a recall of the already available GPT-6 Astra.
• 5:59
  SPEAKER_00: The mathematics story offers a different test of whether a large computing bill bought something checkable.
• 6:05
  SPEAKER_00: OpenAI reports that its agents reached a result on the Navier-Stokes problem after 88 hours, using about 130 billion output tokens on that problem.
• 6:15
  SPEAKER_00: Formalization and verification took another 17 hours.
• 6:19
  SPEAKER_00: Navier-Stokes concerns how fluids behave.
• 6:22
  SPEAKER_00: OpenAI released a written proof and a version in lean, a language that lets software check a formal chain of reasoning.
• 6:29
  SPEAKER_00: The Clay Mathematics Institute has made its own statement.
• 6:33
  SPEAKER_00: On September 11th, it said the problem had apparently been settled, while pointing to its deliberately unhurried process for evaluating the achievement and assigning credit.
• 6:43
  SPEAKER_00: That is stronger evidence than a commentator's impression, but it isn't a final prize ruling.
• 6:49
  SPEAKER_00: The dispute about credit is separate.
• 6:51
  SPEAKER_00: Mathematician Tristan Buckmaster raised concerns about what OpenAI knew of his work.
• 6:57
  SPEAKER_00: OpenAI denies seeing the unpublished work and says its investigation ruled out his recent codex prompts influencing the system.
• 7:05
  SPEAKER_00: Those are competing claims, not something a proof checker can settle.
• 7:09
  SPEAKER_00: Software can check formal reasoning.
• 7:12
  SPEAKER_00: People still have to examine whether it proves the intended statement and how the work should be credited.
• 7:17
  SPEAKER_00: For a leader considering an expensive agent job, decide how you'll judge the answer before authorizing the run.
• 7:24
  SPEAKER_00: Checkability helps justify an experiment.
• 7:27
  SPEAKER_00: It doesn't remove the spending limit or turn every technically valid output into useful work.
• 7:32
  SPEAKER_00: Today's retained auger call concerns Barclays.
• 7:35
  SPEAKER_00: By March 31, 2027, Barclays or Anthropic will publicly state that Claude code use has reached at least 50% of the bank's developers.
• 7:45
  SPEAKER_00: Augur assigns that a 30% probability.
• 7:48
  SPEAKER_00: A hit requires a qualifying press release, results document, or on-record executive statement by the deadline.
• 7:55
  SPEAKER_00: No such statement, or a reported share below half, is a miss.
• 7:59
  SPEAKER_00: This tests public confirmation, not private adoption we can't observe.
• 8:04
  SPEAKER_00: The Google pricing call is withheld for review because Google has already announced higher standard rates.
• 8:10
  SPEAKER_00: It shouldn't be presented as a fresh, unresolved prediction.
• 8:13
  SPEAKER_00: Augur calls have dated hit and miss criteria, so predictions can be checked rather than quietly forgotten.
• 8:20
  SPEAKER_00: They're tracked on the website.
• 8:21
  SPEAKER_00: Adopt the spending check now.
• 8:23
  SPEAKER_00: Pilot expansion where reviewed outcomes support it.
• 8:27
  SPEAKER_00: Watch Argon's wider availability and the formal assessment of the mathematics result without confusing either announcement with a completed decision.
• 8:35
  SPEAKER_00: The written brief with every source is at augerdispatch.ai.
• 8:39
  SPEAKER_00: Portions of the Augur Dispatch are AI generated.
• 8:42
  SPEAKER_00: That is the Augur Dispatch for Friday, October 2nd, 2026.
• 8:47
  SPEAKER_00: Thank you for listening.