Force multiplication IS how concentration happens. If the top 5% of engineers ship 10x the output with AI, the market doesn’t need the next 20%. Their margins collapse, not because they got worse, but because the ceiling moved and dragged the viable floor up with it. Algorithmic trading didn’t replace traders. It made the best ones lethal and hollowed out the middle tier. Same pattern, different domain. “AI makes experts better” and “AI accelerates winner-take-most” are the same sentence.

Three memory regimes are emerging for agents: local, hosted, and shielded remote. Local protects intent but limits capability. Hosted improves capability, but turns strategy into a receipt for providers, policy teams, and courts. Shielded remote is the contested middle ground: frontier inference without full legibility. This will look like a model war. Underneath, it is a market structure fight over who can operate without exposing intent.

Regulation concentrates markets even when it's designed to protect them. The mechanism is simple: compliance costs are mostly fixed. They don't scale with revenue. A firm earning €1M/year and one earning €1B/year pay similar overhead for the same licensing, governance, and reporting obligations. One absorbs it. The other exits. The survivors are selected for capitalization, not quality. This isn't theoretical. In the EU, after GDPR, small IT firms saw a 12% profit drop versus 4.6% for large ones. Market concentration in web services increased 17% within a week of implementation. VC investment in small and micro companies fell by $3.4M per week. Google and Facebook gained market share. The regulation designed to protect users from data exploitation ended up entrenching the largest data exploiters. The EU is now running two more versions of this experiment. MiCA replaces Europe's old crypto registration with a full licensing regime: capital floors up to €150k, application fees, governance mandates, ongoing supervision. Around 130 licenses issued so far. The holders are mostly incumbents: Commerzbank, Société Générale, Coinbase, Crypto dot com. Estimates are that 1,100 to 1,300 firms were active before MiCA. The cost floor is doing its filtering. The AI Act attempts a fix: risk-based tiers and SME carve-outs, a design improvement over MiCA, which has no tiers at all. But tiers sort by risk category, not firm size. A 15-person startup building a hiring algorithm and Google building one face the same compliance bar. Compliance costs for a single high-risk system run €193k to €330k just for the quality management setup. The pattern isn't about bad intent. Each regulation addresses a real problem. The issue is structural: fixed-cost compliance selects for scale as a side effect. Until the cost floor scales with the firm, the selection pressure remains. Tiering by risk is progress. Tiering by firm capacity within each risk level is the unsolved design problem.

There are two ways to centralize a market: Compress margins from above: speed advantages compound, edges get copied, flow consolidates toward whoever controls the pipe. MEV searchers converge on the same strategies until only the fastest extractors survive. Ad auction bots optimize until the bidding floor exceeds what smaller advertisers can sustain. Liquidity agents arbitrage spreads tight enough that manual market makers can't compete. Raise the cost floor from below: make participation expensive enough that only the well-capitalized survive. Licensing regimes that cost more than a small firm's annual revenue. Capital requirements sized for incumbents. Compliance obligations that demand dedicated legal teams before you serve your first customer. Different inputs. Same output. Both are running right now, and they compound.